It’s been another year of relentless AI news - though this time, the headlines felt less like novelty and more like a reckoning. After fears of an “AI plateau” emerged in late 2024, 2025 answered with a barrage of competing models, eye-popping valuations, and the first real signs that artificial intelligence was reshaping jobs, geopolitics, and the tech industry’s power structure.

Subscribe now

Looking back, a few themes dominated: the end of American AI exceptionalism, the scramble for talent and chips, and the growing tension between breakneck innovation and calls for oversight. Some stories played out over months; others landed like bombs. Together, they paint a picture of an industry in flux - simultaneously more capable, more competitive, and more uncertain than ever.

10. DeepSeek

In January, Chinese startup DeepSeek released its R1 reasoning model, demonstrating performance comparable to the best American AI while spending a fraction of the resources.

The release sent shockwaves through Silicon Valley, (arguably) caused Nvidia stock to plummet, and forced a fundamental reassessment of assumptions about AI development requiring massive capital. DeepSeek continued making waves throughout the year with R1-0528 in May, maintaining competitive performance despite export controls.

More than any other single release, DeepSeek changed the narrative: the AI frontier was no longer an American monopoly. More importantly, it challenged the entire premise of the AI arms race - that billions upon billions of capital would be needed to build and run the state of the art models.

9. OpenAI’s Ups and Downs

For better or worse, OpenAI’s story often shapes the narrative of the whole industry - and the company had a rollercoaster of a year. On the upside: a $40B raise at a $300B valuation (the largest private fundraise ever), completing its conversion to a public benefit corporation, and hitting $1B in monthly revenue by August. ChatGPT remains synonymous with “AI” in the minds of many, and it was the most-downloaded app of 2025.

But the year also brought challenges. GPT-5’s August launch was underwhelming, and OpenAI faced real competition for the first time from all directions - DeepSeek, Anthropic, Google, and xAI all took turns at the top of the leaderboards.

Still, the company seems well-positioned heading into 2026, even if its dominance is no longer assured.

8. AI Talent Wars

Competition for AI talent was already fierce, but this year took it to another level. In some cases, talent was being poached: Mark Zuckerberg’s “Superintelligence” lab reportedly offered $100M+ packages, and Tencent paid ~$66M for a single OpenAI researcher. Meta attempted to acqui-hire entire startups - Thinking Machines Lab, Perplexity, Runway - before ultimately freezing hiring in August after months of runaway spending.

In other cases, the talent left to build their own empires: Mira Murati raised $2B at a $10B valuation for Thinking Machines Lab; Yann LeCun departed Meta after clashing with Zuckerberg over LLMs vs. “world models”; Intel’s CTO defected to OpenAI. Even Nvidia got in on the action, “licensing” Groq for $20B in what was effectively an acqui-hire. The war revealed both AI’s perceived value and the industry’s dependence on a surprisingly small pool of elite researchers.

7. Licensing Deals

One of the stories I feel like I haven’t focused on enough this year is the evolution of licensing deals in the space. After years of lawsuits and copyright fights - which, to be clear, are still ongoing - we’ve seen bigger and more diverse licensing deals start to settle the landscape.

On the music side, Suno cut deals with major labels that legitimized AI-generated music in a big way. On the text side, Anthropic reached a multibillion-dollar settlement with book publishers whose work it had trained on, perhaps setting a template for other AI companies to follow. And perhaps most interesting of all, OpenAI and Disney struck a $1B deal that will finally bring legally-approved, AI-generated Disney characters to Sora - a watershed moment for AI in creative industries.

6. AI Comes for the Juniors

What had been theoretical became real in 2025. People have been worried about AI coming for white-collar jobs for two years now, and for the first time, it looks like it might be showing up in the data. To be clear, there’s still a lot of conflicting evidence around this - some studies show productivity gains without job losses, while others paint a grimmer picture.

But I’m most compelled by the research showing that job growth for entry-level programmers has collapsed over the last two years, even as openings for senior roles have continued to grow. The generation entering the workforce is facing a fundamentally different landscape than the one that existed just two years ago.

5. Vibe Coding

It’s only been ten months since Andrej Karpathy coined the term “vibe coding,” and yet it ended up as the Collins English Dictionary Word of the Year. In some sense, not surprising - it’s hard to escape coding models if you’re a programmer these days.

Multi-billion dollar industries have spawned from whole cloth in the past year or so. Cursor hit $1B in ARR by year’s end and turned down repeated acquisition offers. Claude Code has become a major revenue driver for Anthropic.

More importantly, I think coding agents have pretty solidly demonstrated the first real incarnation of the modern “AI agent” - autonomous systems that can take on complex, multi-step tasks with minimal human intervention. There’s some debate over whether the current form and tools of vibe coding will persist, but it’s undeniable that the fundamental task of having an LLM generate code is here to stay.

4. Chip Wars

This isn’t a new story, but it gained new importance in 2025. Biden’s AI Diffusion Rule imposed strict export controls in January; Trump rescinded it in May, then brokered a deal allowing H200 sales to China with a 25% revenue cut to the US government. The geopolitics of AI chips became increasingly tangled with the broader US-China relationship.

China, for its part, is working to wean itself off of its Nvidia dependence. In September, Chinese tech giants were banned from buying Nvidia chips entirely, and elsewhere in the year Huawei unveiled chips that it claimed matched Nvidia’s performance. Whether that’s true remains to be seen - but the direction is clear. The chip war has become inseparable from the AI race itself.

3. Model Mayhem

Despite fears of an “AI plateau” emerging in late 2024, reasoning models got dramatically better in 2025 - and not just from OpenAI. All three leading labs spent the year jockeying for the “state of the art” crown, with leadership changing hands almost monthly.

A few of the models that had the “state of the art” label when they were released this year: DeepSeek R1, Grok 3, Gemini 2.5 Pro, Claude 4 Sonnet, GPT-5, Claude 4.5 Sonnet, GPT-5.1, Gemini 3 Pro, Claude 4.5 Opus, and GPT-5.2.

But more than just the competition, we saw benchmark after benchmark toppled this year. It feels almost like old news at this point, but it’s still so remarkable that these models are now better than the vast majority of human programmers and mathematicians - at least when it comes to coding challenges and math olympiads.

2. Infrastructure Investment

A near-pervasive drumbeat this year was infrastructure spending - it’s hard to pick a single roundup to link to, because nearly every week brought new announcements of multi-billion dollar projects. There’s Project Stargate, with $500B in planned spending, but there’s also Meta’s planned $600B in US infrastructure through 2028, not to mention dozens of “smaller” deals from other tech giants.

Perhaps then, it shouldn’t be surprising that there’s growing backlash to the AI infrastructure buildout. Small towns across America have started rejecting data center proposals over water consumption, noise pollution, and rising energy costs - in some cases facing lawsuits from developers for their trouble.

The year exposed a tension at the heart of the AI boom: the industry’s ambitions require physical resources that communities and markets aren’t always willing to provide.

1. The Bubble Question

Is AI in a bubble? The question haunted 2025.

In March, Alibaba’s Joe Tsai warned of data center overbuilding. Throughout the rest of the year, we continued to see various reports that AI spending was becoming difficult to justify from an ROI perspective - one MIT study found that 95% of GenAI pilots at companies showed little to no financial impact. Even leading AI CEOs like Sam Altman and Sundar Pichai have admitted that valuations have become detached from reality.

There’s been a slight correction - AI stocks shed ~$800B in a single November week. But there’s also the revenue numbers to contend with. This isn’t the dot-com bubble where no real money was changing hands. OpenAI and Anthropic are reaching double-digit billions in ARR, and AI startups like Cursor are growing from zero to a billion in a handful of months.

Yes, there is irrational exuberance. But it’s still wrong to dismiss the entire AI technology wave entirely.

What’s on your top ten list? What stories do you think went underreported this year? Let me know in the comments!

Leave a comment

I’m hoping to do one last post this year, if time allows - a look back at the biggest stories of 2025. But if not: I’m very grateful to all of you reading, and looking forward to what’s to come in 2026!

Empire State Defiance

Last week, New York Governor Kathy Hochul signed the RAISE Act, which requires companies with over $500 million in revenue that develop large AI systems to publish safety protocols, report critical incidents within 72 hours, and face penalties up to $3 million for violations.

Why it matters:

• The law goes further than California’s SB 53, establishing a dedicated enforcement office within the state’s Department of Financial Services. The bill’s sponsor calls it “the strongest AI safety law in the US.”

• The timing is particularly significant - coming just a week after Trump’s executive order attempting to override state AI laws, the law is both a regulatory achievement and a direct challenge to federal preemption efforts.

• Even though the law won’t take effect until January 2027, it may orce AI companies to comply with the strictest standards to operate nationwide - giving New York and California outsized influence over the industry’s future despite Trump administration opposition and the absence of federal rules.

Subscribe now

Elsewhere in AI geopolitics:

• China’s AI regulations require chatbots to pass a 2,000-question ideological test, spawning specialized agencies that help AI companies pass.

• Italy’s antitrust authority orders Meta to suspend WhatsApp terms that bar rival AI chatbots from the platform.

• The Pentagon partners with xAI to embed Grok-based AI systems directly into GenAI.mil as soon as early 2026.

• And tech lobbyists say David Sacks’ push for an executive order blocking state AI laws undercut their efforts to craft a permanent federal solution.

Groq Talk

Nvidia has struck a “licensing” deal with AI chip startup Groq for $20 billion, acquiring rights to use the company’s low-latency chip designs while bringing aboard Groq’s CEO and other executives.

Between the lines:

• This deal represents something of a strategic pivot for Nvidia - rather than competing head-on with every AI chip competitor, the company is selectively absorbing promising technologies to maintain dominance across different market segments.

• Groq’s specialty in ultra-responsive, low-latency chips fills a specific gap in Nvidia’s portfolio, particularly for inference workloads where speed of response matters more than raw training power - a growing market as AI models move from development to deployment.

• This transaction follows a pattern of “acqui-hiring plus licensing“ deals (similar to the Meta-Scale, Google-Character, and Microsoft-Inflection arrangements) that allow big tech to sidestep regulatory scrutiny while still absorbing critical talent.

Elsewhere in AI infrastructure:

• US data centers and the Pentagon face growing national security concerns over their increasing reliance on Chinese batteries.

• US companies sold $1.7T of investment-grade bonds in 2025, nearing the 2020 record, with AI infrastructure borrowing making up approximately 30% according to Goldman Sachs.

• ByteDance has made preliminary plans to spend approximately $23B in AI capex in 2026, up from ~$20B in 2025, with ~$12B budgeted for AI processors.

• Nvidia plans to begin shipping its H200 chips to China before mid-February 2026 with initial shipments expected to be 40,000 to 80,000 units.

• And the global AI boom is being driven by thousands of newcomers diluting Big Tech’s infrastructure dominance, with US data center credit deals hitting $178.5B in 2025.

Alexa, Play Defense

Amazon faces a strategic crossroads as AI shopping agents threaten to disintermediate its customer relationships, forcing the e-commerce giant to choose between blocking these bots or embracing them.

The big picture:

• AI agents promise to revolutionize e-commerce by autonomously searching and purchasing products across the web, potentially generating $1 trillion in US retail revenue by 2030. But this shift threatens Amazon’s margins, as ChatGPT et al would likely require a transaction fee.

• Amazon is currently playing defense - blocking 47 AI bots from crawling its site, suing Perplexity for unauthorized scraping, and developing its own tools like Rufus and Buy For Me. Meanwhile, competitors like Walmart and Shopify are adopting a “frenemy” strategy, partnering with AI companies while setting guardrails.

• Despite the hype, agentic commerce remains nascent with significant technical limitations - only single-item purchases, no loyalty program integration, and frequent glitches - giving Amazon a narrow window to shape the market before these tools mature and consumer behavior shifts permanently away from direct retailer websites.

Elsewhere in the FAANG free-for-all:

• Google DeepMind VP of Research Pushmeet Kohli discusses AlphaFold 2 five years after its launch, covering topics like hallucinations and building an “AI co-scientist”.

• Amazon, Microsoft, and Google pledged a combined $67.5B in Indian investments since October as part of an AI spending frenzy in India, with 80% announced in December.

• Alphabet agrees to acquire data center company Intersect for $4.75B in cash plus existing debt to expand its AI data center footprint.

• Google’s TPUs and Gemini are competitive, but Nvidia’s and OpenAI’s first-mover advantage makes their lead seem durable as Google faces the innovator’s dilemma.

• And Josh Woodward, head of Google Labs, is credited with turning around the Gemini app, growing its MAUs from 350M in March to 650M by October.

Elsewhere in AI anxiety:

• Cursor CEO Michael Truell warns that “vibe coding” advanced projects may create “shaky foundations” that eventually crumble.

• Some US schools are deploying AI surveillance tech like facial recognition and listening devices, though critics say there is little evidence they make schools safer.

• Google’s Nano Banana Pro and OpenAI’s ChatGPT Images can make nonconsensual bikini deepfakes from photos of fully clothed women, prompting Reddit to ban r/ChatGPTJailbreak.

• OpenAI details efforts to secure its ChatGPT Atlas browser against prompt injection attacks, including building an “LLM-based automated attacker.”

• And some experts in human-computer interaction say making AI chatbots act humanlike creates cognitive dissonance for users over how much to trust them.

Things happen

Karpathy’s 2025 LLM year in review. OpenAI rolls out Your Year with ChatGPT, a Spotify Wrapped-like feature. AI can work across scales like steel and the steam engine before it. MiniMax releases M2.1 with enhanced coding capabilities. The NeurIPS transfromation from small academic conference to massive industry event with 24,000+ attendees. Cursor-developer Anysphere acquires Graphite. The AI sector minted 50+ new billionaires in 2025. Chinese AI startup Z.ai releases GLM-4.7 with improved coding performance. Browser extensions with 8M+ installs collected and sold users’ AI chatbot conversations. OpenAI staff discussed prioritizing sponsored content in ChatGPT responses. Local AI is driving the biggest change in laptops in decades. Measuring AI ability to complete long tasks. OpenAI introduces a framework to evaluate chain-of-thought monitorability. Salesforce regrets firing 4000 staff and replacing them with AI.

Last week’s roundup

Flash Forward

Fast (pun intended) on the heels of its flagship Pro model, Google launched Gemini 3 Flash, an impressively high-scoring model given its reduced cost and latency.

The big picture:

• Gemini 3 Flash outperforms its predecessor by significant margins (33.7% vs 11% on Humanity’s Last Exam) while being positioned as a “workhorse model” at $0.50 per million input tokens.

• The Flash model series has become Google’s most popular offering, suggesting developers prioritize deployment economics over cutting-edge performance (though there is a difference between dollars spent and tokens spent).

• Following reports of OpenAI’s internal “Code Red” memo as Google gains consumer market share, both companies are locked in a rapid-fire release cycle that’s processing over 1 trillion tokens per day through Google’s API alone.

Subscribe now

Elsewhere in frontier models:

• Mistral launches Mistral OCR 3, featuring improvements in processing forms, scanned documents, complex tables, and handwriting, priced at $2 per 1,000 pages.

• Nvidia launches Nemotron 3, a family of AI models using a hybrid mixture-of-experts architecture and the Mamba-Transformer design, in 30B, 100B, and ~500B sizes.

• The Allen Institute for AI launches Bolmo 7B and Bolmo 1B, claiming they are “the first fully open byte-level language models“, built using its Olmo 3 models.

• And OpenAI releases GPT‑5.2-Codex, with improvements on long-horizon work through context compaction, stronger performance on large code changes, and more.

Elsewhere in the FAANG free-for-all:

• Google is working on a new initiative to make its AI chips run PyTorch better and is collaborating closely with Meta, as the two discuss Meta using more TPUs.

• Meta is developing a new image and video-focused AI model codenamed Mango, expected to be released in H1 2026 along with its new LLM dubbed Avocado.

• Amazon’s AGI team lead, Rohit Prasad, is leaving at the end of 2025; AWS SVP Peter DeSantis will lead a group combining AI, silicon, and quantum computing teams.

• And Google brings its vibe coding tool Opal to Gemini on the web, letting users create their own custom AI-powered mini apps, called Gems, which can be reused later.

Cracks in the Foundation

A wave of construction delays, funding pullbacks, and market skepticism is hammering AI infrastructure companies, erasing tens of billions in market value and raising urgent questions about whether the frantic race to build data centers can deliver returns that justify the massive capital investments.

Why it matters:

• CoreWeave has lost $33 billion in value over six weeks - a 46% plunge - while data center landlord Fermi’s stock nearly halved after Amazon withdrew $150 million in construction funding, and Oracle pushed back some OpenAI data center completions to 2028.

• The financial stress appears to be mounting across the sector. CoreWeave operates on razor-thin 4% margins while paying double that in interest on its debt - and even profitable giants like Oracle and Broadcom saw double-digit stock declines after disappointing investors on spending timelines.

• As one investor put it, “every single AI company on the planet is saying if you give me more compute I can make more revenue” - but the market is no longer taking that promise on faith.

Elsewhere AI infrastructure:

• A study finds that AI’s 2025 power demand could hit 23GW, exceeding 2024 Bitcoin mining levels, with carbon emissions reaching 32.6M to 79.7M tons compared to NYC’s 50M.

• Hut 8 partners with Fluidstack to build an AI data center in Louisiana for Anthropic, backed by a 15-year, ~$7B lease starting with 245MW of computing capacity.

• Nvidia acquires SchedMD, the developer of Slurm, an open-source AI workload management system, and will continue distributing it on an open-source basis.

• And Oracle signed ~$150B of data center leases in the three months ending November 30, raising its total data center and cloud capacity commitments to $248B.

Elsewhere in AI anxiety:

• The US FTC is investigating Instacart’s AI pricing tool after a study showed different prices for the same items at the same stores, causing CART stock to drop over 8%.

• Hollywood is divided over AI adoption, with some opposing all AI use while directors like Timur Bekmambetov are exploring ways to embrace AI tools.

• Copywriters report that generative AI has led to their work being used for training, layoffs, plummeting wages, and freelancers losing clients.

• And Elizabeth Warren and two other Democratic senators wrote to Google, Microsoft, Amazon, Meta, and others about AI data centers’ role in rising electricity bills.

China’s Secret Chip Lab

China has secretly built a working prototype of an advanced chip-making machine that could eventually break Western dominance in semiconductor manufacturing, years ahead of Western estimates.

Between the lines:

• The prototype, completed in early 2025 in a high-security Shenzhen lab, is operational and generating EUV light but hasn’t produced working chips yet. China aims for 2028, though experts say 2030 is more realistic.

• This is China’s “Manhattan Project” for semiconductors, coordinated by Huawei and involving thousands of engineers working under strict secrecy. Some recruits were given fake IDs to conceal their identities, and the effort involves salvaging parts from older machines and exploiting secondary markets to bypass export controls.

• Yet despite this breakthrough and massive AI investment battles between the U.S. and China, recent analysis reveals China’s AI chip capabilities are falling further behind, not catching up - Nvidia’s chips are currently five times more powerful than Huawei’s best offerings, expanding to seventeen times by 2027.

• On the other hand, America’s concentrated $350+ billion bet on AI contrasts with China’s diversified strategy spreading investments across electric vehicles, batteries, solar panels, and other manufacturing sectors - raising the prospect that the US could win the AI battle but lose the broader economic war.

Elsewhere in Chinese AI:

• At least six Chinese AI companies, including chip firms Montage and GigaDevice, are set to IPO in Hong Kong in the coming weeks, aiming to raise up to $1B.

• Chinese AI chipmaker MetaX’s shares jumped as much as 755% in its Shanghai debut, after a heavily oversubscribed ~$585.8M IPO, pushing its market cap past ~$42B.

• Nvidia told its Chinese clients that it is evaluating adding production capacity for its H200 chips after orders exceeded its current output level.

• And Taiwan opened its largest AI supercomputing data center, with Nvidia’s Blackwell chips, in a major effort for its push for sovereign AI and chip industry innovation.

Elsewhere in AI geopolitics:

• The UK AI Security Institute reports that AI models are rapidly improving at potentially dangerous biological and chemical tasks and show fast jumps in self-replication capabilities.

• a16z released a nine-pillar AI policy framework for Congress, including restricting AI use for under-13s without parental consent and keeping a role for states.

• A look at how David Sacks and AI adviser Sriram Krishnan overcame MAGA opposition to Trump’s AI executive order by talking to lawmakers and modifying its language.

• And Trump’s AI executive order directs federal agencies to look into withholding federal funding from states passing “the most onerous and excessive laws” around AI.

Skills Without Borders

A lightweight “skills” format introduced by Anthropic in October for teaching AI assistants repeatable workflows is rapidly becoming an industry standard, with OpenAI implementing it within ChatGPT and Codex - signaling a rare moment of interoperability in the competitive AI landscape.

The big picture:

• Skills represent a deceptively simple but powerful specification: just a folder containing a Markdown file with instructions and optional resources that any AI system with filesystem access can implement. They’re more accessible than complex agent frameworks, more portable than platform-specific features, and more standardized than custom prompts.

• Anthropic is betting on an ecosystem play, launching a partner directory with pre-built skills from Notion, Canva, Figma, and Atlassian while adding enterprise management features that let admins provision skills organization-wide.

• The speed of OpenAI’s adoption (appearing just weeks after Anthropic’s October launch) and the emergence of community repositories suggest skills are solving a real pain point. The question now is whether this informal standard will be formalized - perhaps through initiatives like the new Agentic AI Foundation - or remain an organic speci that evolves through practice rather than governance.

Elsewhere in OpenAI:

• OpenAI has held preliminary discussions with investors about raising tens of billions of dollars, potentially as much as $100B, at a valuation of around $750B.

• OpenAI hires former UK chancellor George Osborne to lead OpenAI for Countries, the global expansion arm of its $500B Stargate data center initiative.

• OpenAI opens app submissions for review and publication in ChatGPT, where users can discover or search for apps in a new app directory.

• OpenAI debuts FrontierScience, a benchmark to measure models’ expert-level scientific capabilities with 700+ questions, finding GPT-5.2 is its strongest model.

• And OpenAI updates ChatGPT Images to generate images up to 4x faster and edit uploads more precisely, adding a new section in ChatGPT’s app and website for images.

Things happen

Tencent names former OpenAI researcher Yao Shunyu as chief AI scientist. xAI’s lack of enterprise sales experience is holding back potential customers. AI image generators now mimic phone camera traits to avoid the uncanny valley. Q&A with Sam Altman on OpenAI’s “code red” call, enterprise strategy, and IPO plans. Claude ran a vending machine and lost $1,000+ after dropping prices to zero. Mira Murati’s lab makes Tinker generally available. YouTube terminates two channels that used AI to create fake movie trailers. Google launches CC, an AI assistant that delivers personalized daily briefings. Mark Zuckerberg’s all-in AI bet includes suffocating micromanagement, sources say. Google removes AI videos of Disney characters after cease-and-desist letter. Amazon rolls out Ask This Book to Kindle; authors can’t opt out. Twenty-four companies join the US Genesis Mission to boost AI for scientific discovery. The AI boom is delaying municipal projects as workers shift to data centers. Anthropic exec forces AI chatbot on gay Discord community, members flee. Chinese AI infrastructure companies surge despite Trump’s tariffs. Trump’s AI Executive Order could invalidate GOP state laws targeting “Big Tech censorship.” GOP governors push ahead with their own AI bills despite Trump’s EO.

Last week’s roundup

GPT-5.2

OpenAI has launched GPT-5.2, its most capable AI model yet, in the wake of the company’s internal “code red.” The new model series (comprising Instant, Thinking, and Pro variants) delivers substantial performance gains across professional tasks, particularly in knowledge work benchmarks.

Why it matters:

• GPT-5.2 targets professional workflows and AI agents specifically, with claims of better performance in spreadsheets, presentations, code, and “complex, multi-step projects” - plus reduced hallucinations in its Thinking model to build trust for enterprise use.

• While the capabilities are impressive, this release also signals a future where the path forward is solving real-world professional tasks rather than chasing abstract intelligence metrics - Anthropic has previously done the same, declaring that it would focus on real-world coding capability rather than benchmarks.

• Ultimately, models like this take months to train and launch, but CEO Fidji Simo’s acknowledgment that “code red” resources were “helpful” is an acknowledgment that the landscape has indeed changed - OpenAI is no longer solely setting the pace for model progress.

Subscribe now

Elsewhere in frontier models:

• Runway launches GWM-1, its first world model that uses frame-by-frame prediction to simulate physics, and updates Gen 4.5 to add native audio.

• Mistral releases Devstral 2, a 123B-parameter AI coding model requiring at least four H100 GPUs, alongside Devstral Small, a 24B-parameter model for local use.

• Chinese AI startup Z.ai debuts GLM-4.6V, open-weight vision models with native function calling available in 106B- and 9B-parameter versions.

• And Anthropic, OpenAI, Block, Google, AWS, Microsoft, and others launch the Agentic AI Foundation to build open-source agent standards under the Linux Foundation.

Federal Override

The Trump administration is aggressively reshaping AI policy with a dual strategy: domestically restricting states’ ability to regulate artificial intelligence companies while internationally allowing tech giants like Nvidia to sell advanced chips to China in exchange for a 25% revenue cut to the US government.

Between the lines:

• President Trump signed an executive order threatening to withhold $42 billion in broadband funding from states with AI regulations his administration deems too restrictive, as well as establishing an “AI Litigation Task Force” under Attorney General Pam Bondi to actively challenge state AI laws.

• The order targets states like Colorado (anti-discrimination provisions) and California (catastrophic risk mitigation requirements), framing consumer protections as “ideological bias” that could stifle innovation and America’s competitiveness against China.

• President Trump also announced that Nvidia can sell its H200 AI chips to “approved customers” in China, but only if the US government receives a 25% cut of the revenue. The H200s are something of a middle ground - they’re nearly six times more powerful than chips currently allowed in China but significantly behind the Blackwell chips US companies are using.

• The gamble is that the US can maintain its AI lead through a deregulated domestic market and controlled chip exports that generate revenue, but critics see a “lawless Wild West” at home and a “colossal national security failure” abroad that could ultimately strengthen China’s military and indigenous chip industry.

Elsewhere in AI geopolitics:

• New US OMB guidance states that LLMs procured by federal agencies must comply with two “unbiased AI principles“: “truth-seeking” and “ideological neutrality”.

• The NY governor is reportedly considering a rewrite of the RAISE Act, the AI bill that recently passed the NY legislature, with text copied verbatim from California’s SB 53, after signing bills requiring ads to disclose AI-generated performers and mandating consent of heirs to use a deceased person’s likeness commercially.

• US state AGs sent a letter to Meta, Microsoft, Google, Apple, and nine others, warning that their chatbots’ “delusional outputs” could be violating state laws.

• The US DOJ detains two men for allegedly violating export controls by trying to smuggle $160M+ of Nvidia chips to China; a third man pleads guilty.

• And the US DOD says it has chosen Google’s Gemini for Government to power its new GenAI.mil platform for the US military, as part of a $200M contract from July.

Elsewhere in AI infrastructure:

• Broadcom CEO Hock Tan reveals that Anthropic placed a $10B order for Google’s Ironwood TPU racks in Q3 and an additional $11B order in Q4.

• Blue Origin has worked for over a year on tech for orbital AI data centers, while SpaceX plans to use upgraded Starlink satellites for AI computing payloads.

• Nvidia says it hasn’t seen substantiation of chip smuggling via data centers outside of China after The Information’s DeepSeek story.

• Big Tech’s AI data center plans face a power crunch that could deflate the AI “bubble” as they need an estimated 44GW of additional capacity by 2028.

• And Amazon and Microsoft plan to invest $35B+ in India and $17.5B in India respectively, over the next four years on AI and cloud computing.

Creative Licensing

Disney and OpenAI have struck a $1 billion deal that will allow Sora users to create and share short videos featuring over 200 Disney, Marvel, Pixar, and Star Wars characters - marking the first major licensing partnership between a legacy entertainment giant and a generative AI company.

Between the lines:

• The three-year agreement positions Disney as both a content licensor and major OpenAI customer, with plans to integrate ChatGPT across its operations and use OpenAI’s APIs to build new Disney+ experiences, while curated fan-made Sora videos will stream on the platform starting in early 2026.

• The deal’s carefully defined boundaries reveal Hollywood’s red lines: no talent likenesses, no character voices, and OpenAI must allow Disney to “set and evolve guardrails over time.” Disney CEO Bob Iger is framing this as inevitable adaptation rather than capitulation, arguing that “no human generation has ever stood in the way of technological advance.”

• Ultimately, this is as much about strategic positioning in the AI race as it is about licensing revenue - Disney also this week sent a cease-and-desist letter to Google, accusing the tech giant of “massive scale” copyright infringement by training AI models on Disney’s characters - following similar actions against Meta, Character.AI, Midjourney, and Minimax.

Elsewhere in OpenAI:

• Sam Altman reflects on OpenAI’s 10th anniversary, highlighting a decade of breakthroughs and the path toward AGI that benefits all of humanity.

• Fidji Simo says ChatGPT’s “adult mode” is expected to launch in Q1 2026 once the company improves its age prediction capabilities.

• OpenAI and Instacart launched a grocery shopping experience inside ChatGPT that lets customers brainstorm meal ideas and check out directly.

• The company warns that its upcoming frontier AI models are likely to pose a “high” cybersecurity risk as their cyber capabilities accelerate.

• Slack CEO Denise Dresser is leaving the company to join OpenAI as its Chief Revenue Officer after serving in the role since November 2023.

• And Sam Altman said OpenAI plans to end its “code red” after releasing a model in January 2026 with improved image generation, speed, and personality.

Elsewhere in AI anxiety:

• Stanford researchers developed AI hacking bot Artemis, which surpassed nine out of 10 penetration testers by rapidly finding bugs in the university’s network.

• A study found 15 TikTok accounts posting AI videos of sexualized underage girls have nearly 300K followers, with TikTok saying that 14 of the accounts don’t violate its rules.

• The International Committee of the Red Cross warned that AI models are fabricating research papers, journals, and archives.

• The New York Times sued Perplexity, claiming the AI startup violated its copyrights and kept using its content despite repeated demands over the past 18 months.

• A developer who accidentally found CSAM in AI training data was banned by Google for reporting it.

• And McDonald’s Netherlands pulled an AI-generated Christmas ad after social media backlash.

Google has a lot going on

The headline this week was GPT-5.2, but Google is keeping up with its nonstop launch cadence.

What’s the latest:

• Disco, a Gemini 3-powered experiment that suggests and creates GenTabs, web apps based on the user’s open browser tabs and Gemini chat history.

• An enhanced Gemini Deep Research agent accessible to developers via its new Interactions API, along with a new DeepSearchQA benchmark.

• Fully managed, remote MCP servers to help developers connect AI agents to services such as Maps, BigQuery, Compute Engine, and Kubernetes Engine.

• Gemini 3 Pro sets new vision AI benchmark records, including in complex visual reasoning, beating Claude Opus 4.5 and GPT-5.1 in some categories.

• Amin Vahdat, who leads the company’s AI and infrastructure team, to chief technologist for AI infrastructure, reporting to Sundar Pichai.

• A cheaper AI Plus plan in India, costing ~$2.21 per month for the first six months and ~$4.44 thereafter, to compete with ChatGPT Go.

• Google has told advertisers that it plans to bring ads to Gemini in 2026; Google’s VP of Global Ads says there are no plans for ads in the Gemini app.

• DeepMind plans to open its first “automated science laboratory” in the UK in 2026, focused on using AI tools to develop new materials for chips and more.

• The company details the steps it is taking to secure Chrome’s upcoming agentic browsing features, like a “User Alignment Critic” model that vets each AI agent action.

• And Google details its plans for two different categories of AI-powered smart glasses coming as soon as 2026: one with in-lens displays and one that’s audio-focused.

Elsewhere in the FAANG free-for-all:

• Meta’s new AI model, codenamed Avocado, may launch in spring 2026 as a “closed” model trained using Google’s Gemma, OpenAI’s gpt-oss, and Qwen.

• Tensions have emerged between Meta’s TBD Lab and longtime lieutenants to Mark Zuckerberg over computing resources, company goals, and priorities.

• Microsoft Research, Providence, and UW developed the GigaTIME AI model, which can analyze tumors in a fraction of the time and cost of existing methods.

• And Meta acquires Limitless, which makes a pendant-style AI wearable that records and transcribes real-world conversations.

Things happen

xAI partners with El Salvador to deploy Grok across 5,000+ public schools. Time names “the Architects of AI“ as its 2025 Person of the Year. Cursor launches Visual Editor, a vibe-coding product for designers. How the Model Context Protocol became an industry standard. An overview of AI in 2025, including arguments for and against above-trend capabilities growth. AI browsers are forcing web developers to rethink whether they’re designing for humans or crawlers. The US FDA qualifies the first AI drug development tool to help assess a fatty liver disease. How Chinese companies hire Kenyan AI annotators via opaque middleman networks to avoid accountability. Inside the creation of AI actress Tilly Norwood, which took 2,000 iterations. Several companies are “refounding” to add AI features. Reasoning models now represent over half of all usage, an analysis of 100T+ tokens shows. AGI may never emerge because the concept ignores physical realities of computation. ChatGPT was 2025’s most downloaded free app in the US App Store.

Last week’s roundup

OpenAI’s Code Red

OpenAI CEO Sam Altman has declared a “code red,” ordering staff to refocus on improving ChatGPT as the company’s once-dominant position faces its most serious challenges yet.

Between the lines:

• OpenAI is reportedly shelving ambitious initiatives like ads, shopping agents, and a personal assistant called Pulse to focus on core ChatGPT improvements - with daily calls and temporary team transfers to accelerate development.

• It’s a striking role reversal from three years ago - when Google declared its own “code red” in response to ChatGPT’s explosive debut. Google has orchestrated a remarkable turnaround, with Gemini 3 outperforming GPT-5 on key benchmarks, its mobile app surging to 650 million monthly users, and its custom TPU chips allowing it to train models without relying on expensive Nvidia hardware.

• Ultimately, OpenAI’s aggressive expansion into multiple product lines simultaneously - from video generation (Sora) to coding tools to planned advertising and shopping features - may have stretched the company too thin.

Subscribe now

Elsewhere in frontier models:

• Mistral launches Mistral 3, a family of 10 models under the Apache 2.0 license, including its new flagship Mistral Large 3 and nine smaller Ministral 3 models.

• Chinese short-video company Kuaishou launches Kling Video O1, saying it is the first multimodal AI model to unify video generation, editing, and post-production.

• DeepSeek releases DeepSeek-V3.2 and DeepSeek-V3.2-Speciale, which it calls “reasoning-first models built for agents”, after releasing V3.2-Exp in September.

• Runway debuts the Gen-4.5 AI video generation model, highlighting its physical accuracy and holding the top spot on Artificial Analysis’ Video Arena leaderboard.

• Nvidia announces Alpamayo-R1, an AI model for autonomous driving research, calling it the “first industry-scale open reasoning vision language action model”.

• Prime Intellect debuts INTELLECT-3, a 106B open source MOE model it claims outperforms larger models.

• Google rolls out Gemini 3 Deep Think to Google AI Ultra subscribers in the Gemini app after delays for safety evaluations.

• And Apple releases Starflow-V, an open weights video model.

Elsewhere in OpenAI:

• OpenAI is testing training LLMs to produce “confessions”, or self-report how they carried out a task and own up to bad behavior like lying or cheating.

• OpenAI’s nonprofit foundation plans to award $40.5M in grants in 2025 to 208 US nonprofits, up from only $7.5M donated in 2024.

• OpenAI takes a stake in Thrive Capital’s Thrive Holdings and says it will embed agents in its companies, which already include an accounting and an IT business.

• OpenAI’s data center partners are set to rack up nearly $100B in debt, with banks potentially lending another $38B to Oracle and Vantage to build more OpenAI sites.

• And a US federal judge rules that OpenAI must produce 20M anonymized ChatGPT chat logs in the copyright lawsuit brought by The New York Times and other news outlets.

Anthropic’s IPO

The AI IPO race is intensifying as Anthropic (and OpenAI) appear to be preparing for public listings at unprecedented valuations, despite incurring massive losses and facing unpredictable financial forecasts.

What to watch:

• This week, Anthropic made concrete moves toward going public - hiring Wilson Sonsini (the firm behind Google, LinkedIn, and Lyft IPOs) and bringing on Krishna Rao as CFO, who led Airbnb’s successful public debut.

• The company is valued at $183 billion, is on track to hit $10 billion in annualized revenue - more than 10x its 2024 figure - and just announced its first-ever acquisition of Bun, a high-performance JavaScript runtime, as its Claude Code product hits $1 billion in run-rate revenue.

• Anthropic’s success challenges the narrative that the AI race requires unprecedented capital expenditures: the company generates 2.1 times more revenue per dollar of computing costs compared with OpenAI.

• But both IPOs will be a crucial test of whether public markets will embrace the expensive, unprofitable AI research labs that have thrived in private markets, potentially setting the tone for the entire AI industry’s financial future and sustainability.

Elsewhere in Anthropic:

• Snowflake and Anthropic announced a multiyear $200M deal to make Claude models available on Snowflake’s platform and deploy AI agents for enterprises.

• Anthropic says 60%+ of its business customers use more than one Claude product, a trend it began noticing in summer 2025 after Claude Code’s rise in popularity.

• Anthropic’s employees self-report using Claude in 60% of their work, achieving a 50% productivity boost, mostly for debugging and code understanding.

• Anthropic’s societal impacts team studies AI’s broad societal risks to tackle “inconvenient truths”, beyond typical safety teams at AI startups.

• A look at the exfiltrated 14K-token document that Claude calls its “Soul overview”, confirmed to be valid by Anthropic employees.

• And using the SCONE-bench benchmark of 405 blockchain smart contracts, Claude Opus 4.5, Claude Sonnet 4.5, and GPT-5 developed exploits together worth $4.6M.

Elsewhere in AI infrastructure:

• Jensen Huang says Nvidia doesn’t know if China would accept its H200 AI chips should the US relax export controls on them, after a meeting with President Trump.

• An in-depth look at how the latest Google TPU generation positions Google as the most threatening challenger to Nvidia’s AI chip dominance.

• The AI boom has led to high demand and more pay for the construction workers who build data centers, with a trade group estimating there’s a shortage of ~439K workers.

• And Dario Amodei suggests some AI companies are taking on too much risk by “YOLO-ing” and committing to spend hundreds of billions on data centers.

Three-Year Mark

Three years after ChatGPT’s release sparked a global AI frenzy, Americans remain deeply divided over the technology - with the US showing more fear and less excitement about AI than nearly any other surveyed nation, even as it remakes the world around us.

The big picture:

• For starters, artificial intelligence is becoming one of the most popular college majors in the U.S., with MIT’s AI program now the second-largest major on campus and thousands of students enrolling in new AI degrees across the country.

• Likewise, AI mania has fundamentally reshaped Wall Street, driving a 64% surge in the S&P 500 while creating unprecedented market concentration around a handful of tech giants. However, this success comes with a caveat: the seven largest companies now account for 35% of the S&P 500, creating significant concentration risk.

• And while legitimate concerns exist (electricity usage, carbon emissions), the technology’s genuine utility - as a knowledge assistant, translator, and productivity tool - tends to go unappreciated.

Elsewhere in AI geopolitics:

• President Trump’s plans to block state-level AI regulations via a preemption proposal face opposition from Democrats, Republicans, and consumer groups.

• The EU expects to launch a formal bidding process for its AI gigafactories in early 2026 and close it in summer 2026, as the bloc seeks to catch up with the US.

• Jimmy Wales says Wikimedia is working with Big Tech on AI deals similar to Google’s and reader donations are “not to subsidize OpenAI costing us a ton of money”.

• The European Commission opens a new antitrust investigation into Meta over its rollout of its AI assistant in WhatsApp; the probe will not fall under the DMA.

• David Sacks’ AI and crypto policies in Trump’s White House benefit his investments, those of his Silicon Valley friends, and the All-In podcast he co-hosts.

• And scientists are (once again) becoming increasingly worried AI will sway elections.

Elsewhere in AI anxiety:

• Some people are feeding years of medical records into chatbots like ChatGPT, despite privacy risks, and receiving generalized or inaccurate diagnoses in response.

• YouTube creators are using AI tools to make videos for kids and babies, raising concerns that such AI content may negatively impact early brain development.

• AI Forensics found that 354 AI-focused TikTok accounts pushed 43K posts made with GenAI tools that hit 4.5B views, including posts with anti-immigrant and sexual material.

• Researchers unveiled PropensityBench, a benchmark showing how stressors like shortened deadlines increase misbehavior in agentic AI models during task completion.

• Pangram Labs found that ~21% of the 75,800 peer reviews submitted for ICLR 2026, a major ML conference, were fully AI-generated, and 50%+ contained signs of AI use.

• And Unit 42 details how underground hacking forums advertise and sell custom, jailbroken, and open-source AI hacking tools such as WormGPT and KawaiiGPT.

re:Invent 2025

Amazon held its AWS re:Invent 2025 conference this week, and we got a fairly impressive set of new launches, including new models, chips, and agents.

What to watch:

• Amazon Nova, a second-generation AI model family including Nova Lite, Nova Pro, Nova Sonic, and fully multimodal reasoning model Nova Omni, released to limited customers.

• Trainium3, an AI chip that is 4x faster than Trainium2 and can cut AI training and operating costs by up to 50% compared to equivalent GPUs.

• Trainium3 UltraServer, a system powered by its 3nm Trainium3 AI training chip, alongside a tease of Trainium4, which will work with Nvidia’s chips.

• Nova Forge, a $100,000/year service allowing clients to customize Amazon’s AI models at various stages of training and refine open-weight models.

• Three frontier agents: Kiro autonomous agent, AWS Security Agent, and AWS DevOps Agent, each focused on different aspects of software development.

• Bedrock AgentCore expansion with new tools for managing agent boundaries, agent memory capabilities, and agent evaluation features.

• DevOps Agent, an AI-enabled tool designed to help clients quickly identify root causes of outages and implement fixes, available in preview.

• AWS AI Factories, which lets customers deploy AWS infrastructure, including Trainium chips and Nvidia GPUs, in their existing data centers.

• And an interview with AWS CEO Matt Garman on his AI vision, extending Amazon’s cloud market lead, adding AI to AWS services, and AI efficiencies.

Things happen

ByteDance launches an AI voice assistant powered by its Doubao LLM. ChatGPT told a violent stalker to embrace the “haters”. Thoughts on AI progress and model limitations. A profile of Cursor CEO Michael Truell, 25. Google threatens both OpenAI and Nvidia. Stanford tool downranks antagonistic posts on X. US startups adopt open-weight Chinese AI models. Apple’s AI chief is retiring. Everyone in Seattle hates AI. Reverse engineering a legal AI tool exposed 100K+ confidential files. A new AI winter is coming? We gave 5 LLMs $100K to trade stocks for 8 months. Anti-patterns while working with LLMs. Codex, Opus, Gemini try to build Counter Strike. Writing a good Claude.md. Search tool for content created before ChatGPT. CA ballot measure to undo nonprofit conversions, targeting OpenAI.

Last week’s roundup

Opus Day

Anthropic released Claude Opus 4.5 on Monday, reclaiming the crown for best coding model with an 80.9% score on SWE-Bench Verified. The release caps a frenetic week of AI lab competition, coming just days after OpenAI’s GPT-5.1-Codex-Max and Google’s Gemini 3 Pro.

What to watch:

• The new model is dramatically cheaper than the previous Opus release ($5/$25 per million tokens, as opposed to the previous $15/$75), meaning it may actually be cost-effective to use Opus widely as a base model - especially considering it matches its predecessor Sonnet 4.5’s performance while using 76% fewer tokens.

• With this release, Anthropic finally restores its three-tier model hierarchy after Sonnet 4.5 had been outperforming the older Opus 4.1, giving developers clear choices across price-performance curves rather than forcing everyone to upgrade to the flagship model.

• For developers, Anthropic is also releasing three new features to help Claude work with massive tool libraries: Tool Search Tool (for discovering tools on demand), Programmatic Tool Calling (for executing tools through code), and Tool Use Examples (for demonstrating correct usage patterns).

• But all these rapid-fire releases expose a growing evaluation problem: Even Simon Willison struggled to distinguish Opus 4.5’s real-world capabilities from its predecessor Sonnet 4.5, noting he “kept on working at the same pace“ after switching back - suggesting frontier models may be converging faster than our ability to meaningfully differentiate them.

Subscribe now

Elsewhere in frontier models:

• Microsoft unveils Fara-7B, its first agentic small language model designed for computer use, available as an experimental release on Hugging Face and Microsoft Foundry.

• An analysis of Gemini 3’s model card and safety framework report finds the model is excellent but the safety report withholds or makes it difficult to understand key information.

• Ilya Sutskever discusses model jaggedness, why we are moving beyond the “age of scaling,” and SSI’s plan to straight-shot superintelligence in a new Q&A.

• OpenAI merges ChatGPT’s voice mode directly into the main text chat interface by default, though users can still switch back to the original separate voice mode.

• And OpenAI unveils a free shopping research feature in ChatGPT that delivers a personalized buyer’s guide, powered by a custom version of GPT-5 mini.

Short-term memory loss

A severe memory chip shortage, driven by surging demand for AI infrastructure, is threatening to disrupt supply chains across the tech industry. Major manufacturers, including Dell and HP, are warning of unprecedented cost increases that will inevitably be passed on to consumers.

The big picture:

• The crunch stems from manufacturers prioritizing production of advanced, high-bandwidth memory for profitable AI systems over conventional memory chips used in everyday devices - from smartphones and PCs to medical equipment and cars.

• Some companies, like Apple, are leveraging their scale and long-term contracts to secure supply and potentially gain market share. Others, like Lenovo, have begun stockpiling memory inventories up to 50% higher than normal levels, as memory chip prices are forecast to rise 50% through mid-2026.

• The shortage is becoming a bottleneck for the broader tech ecosystem: Logic chip suppliers may see reduced orders if customers can’t secure accompanying memory, while U.S. sanctions limiting Chinese chipmakers’ capabilities are exacerbating supply constraints just as panic buying intensifies the crisis.

• The timing creates a frustrating moment for consumers - GPU prices are finally normalizing after years of volatility, making graphics cards affordable just as the cost of building or upgrading a PC has become prohibitively expensive due to memory shortages.

Elsewhere in AI infrastructure:

• Google’s head of AI infrastructure says the company must double AI compute capacity every six months to meet demand.

• Meta enters electricity trading to help accelerate the construction of new US power plants critical for its AI infrastructure buildout.

• AWS announces a commitment to invest up to $50B to build AI and HPC infrastructure for the US government, starting in 2026 and adding nearly 1.3 GW of capacity.

• Foxconn Chair Young Liu says the company will spend $2B to $3B per year in AI in the next three to five years and is discussing potential investments with Japan.

• And since 2023, data center power demands have delayed 15 coal plants’ retirements as the Trump administration has ordered two power plants to remain open.

Chip shot

Google is intensifying its challenge to Nvidia’s AI chip dominance by pitching major customers on using its Tensor Processing Units (TPUs) in their own data centers, with Meta reportedly considering a multi-billion-dollar deal for 2027.

Between the lines:

• The move prompted Nvidia’s stock to fall 3% and drew a swift public response from the chip giant, which declared its technology remains “a generation ahead” of alternatives like Google’s TPUs.

• It’s also throwing its war chest around - Nvidia CEO Jensen Huang has been proactively countering Google’s moves by investing billions in Anthropic and OpenAI after they explored TPU deals, essentially using capital to maintain GPU lock-in even as Google offers cheaper alternatives.

• Google’s recent technical progress with Gemini 3 and seven generations of TPU development may finally be paying off - Meta is reportedly interested in using TPUs for model training, not just inference, which many had assumed was Nvidia’s unassailable stronghold.

Elsewhere in the FAANG free-for-all:

• Google is starting to bridge OpenAI’s product moat with Gemini’s “dynamic view” option, which converts text answers into interactive, visual outputs.

• A job listing shows Google is developing a new Android-based “Aluminium OS” that is “built with AI at the core”, potentially as a ChromeOS replacement for PCs.

• Meta plans to test an AI-powered personalized daily briefing, designed to compete with ChatGPT’s Pulse, with some Facebook users in NYC and SF.

• Google says ads that some users are seeing in AI Mode are part of a test; the ads have a “sponsored” label and appear at the bottom of the page.

• And Amazon unveils its Autonomous Threat Analysis system, born from a 2024 internal hackathon, to use AI agents competing in teams to identify vulnerabilities.

Open season

China has surpassed the US in downloads of “open” AI models on Hugging Face for the first time, capturing 17% market share versus America’s 15.8%, as Beijing encourages its tech companies to release freely accessible models while US giants like OpenAI and Google maintain closed, proprietary systems.

The big picture:

• Led by DeepSeek and Alibaba’s Qwen models, Chinese AI developers are releasing models weekly at dramatically lower costs - while American tech leaders like OpenAI, Google, and Anthropic maintain tight control over their “closed” proprietary technology to protect lucrative subscription businesses.

• Likewise, foreign investors are warming to China’s AI sector precisely because it lacks the bubble characteristics plaguing US tech, with venture capital firms raising hundreds of millions in recent months to back Chinese startups at valuations one-quarter of American peers.

• But China’s open-source dominance comes with ideological strings attached: researchers have documented that Chinese models embed Communist Party biases and refuse to generate information on sensitive topics like Taiwan or Tiananmen Square, meaning widespread adoption could shape global information flows.

Elsewhere in AI geopolitics:

• Italy’s competition regulator is scrutinizing a Meta policy that excludes rival AI chatbots from WhatsApp, broadening the scope of a probe started in July.

• The US House Homeland Security Committee asks Dario Amodei to testify at a December 17 hearing on how Chinese state actors used Claude Code for cyber-espionage.

• A new network of super PACs plans to raise ~$50M to counter the Leading the Future super PAC and back candidates who prioritize AI regulations.

• Trump signs an EO establishing the Genesis Mission to boost AI innovation, including by using federal scientific datasets to train models and create AI agents.

• New York’s RAISE Act would require AI companies to publish safety protocols and disclose serious incidents, as its co-sponsor is targeted by a pro-AI super PAC.

• And the UK announces a £100M plan to support local AI hardware startups via guaranteed “first customer” payments, estimating its AI market to be worth over £72B.

Elsewhere in AI anxiety:

• An MIT study finds that AI can replace 11.7% of the US labor market, or ~$1.2T in wages, based on the “Iceberg Index” which measures job automation potential.

• Anthropic finds that LLMs trained to “reward hack” by cheating on coding tasks show even more misaligned behavior, including sabotaging AI-safety research.

• Major insurers have recently sought permission from US regulators to offer policies excluding liabilities tied to businesses deploying AI chatbots and agents.

• While the AI bubble feels like the internet bubble of 1999, it may actually be larger and scarier with an unstable US economy and greater exposure for Big Tech.

• Researchers say Russia-aligned Pravda network is engaging in “LLM grooming”, flooding the internet with disinformation to influence chatbots like ChatGPT.

• And interviews with current and former OpenAI employees detail how updates that made ChatGPT more appealing to boost growth sent some users into delusional spirals.

Things happen

Epic games CEO says it makes no sense to label AI-made games since AI will be in everything. Report: Suno spent $32M on compute and $2K on data. 73% of AI startups are just prompt engineering. PopEVE predicts disease-causing mutations better than AlphaMissense. Indian startups are building LLMs for low-resource languages from scratch.Amazon begins previewing Leo satellite internet ahead of 2026 launch. Sam Altman and Jony Ive say their OpenAI device could arrive in less than two years. The bitter lesson of LLM extensions. The USPTO clarifies that AI cannot be an inventor. Character.AI is cutting off ongoing chats for users under 18. Figure AI’s former head of product safety sues the company for wrongful termination. An AI agent bypasses bot detection on online surveys nearly flawlessly. Amazon pushes engineers to use Kiro over Cursor. The humanoid robotics industry relies heavily on hype. Food bloggers warn AI Overviews are burying real recipes. A Mixpanel breach exposed OpenAI API data; OpenAI terminated its use. I don’t care how well your AI works. Replace your boss before they replace you. Karpathy on the implications of AI to schools. PRC elites voice AI-skepticism. Altman: Google’s AI progress could create economic headwinds for OpenAI. The AI workers tell their friends and family to stay away from AI. We’re losing our voice to LLMs.

Last week’s roundup

Gemini 3

Google launched Gemini 3 Pro on Tuesday, its most capable AI model to date. The model, which has generated considerable coverage (including here), demonstrates strong planning, coding, and judgment skills, showing that AI models have progressed beyond hallucinations to subtle, human-like errors.

What to watch:

• Google also released “Nano Banana Pro,” its viral-named Gemini 3-powered image model that excels at following instructions and creates full infographics from short prompts.

• Google debuted Antigravity, an “agent-first” coding tool that leverages Gemini 3 Pro and third-party models, in public preview for Windows, macOS, and Linux.

• Google priced Gemini 3 Pro at $2 to $4 per 1M input tokens and $12 to $18 per 1M output tokens, cheaper than Claude Sonnet 4.5 but more expensive than GPT-5.1.

• Google details a new generative UI implementation that lets Gemini 3 Pro create interactive interfaces for any prompt, launching in the Gemini app and AI Mode.

• And Google plans to release Gemini 3 Deep Think to Google AI Ultra subscribers in the coming weeks, once it passes further rounds of safety testing.

Elsewhere in frontier models:

• OpenAI unveils GPT-5.1-Codex-Max, saying it is “significantly better” at “long-horizon reasoning” and is the first model it has trained for Windows environments.

• xAI’s Grok 4.1 tops benchmarks in emotional intelligence, while its model card also shows a marked increase in sycophancy compared to Grok 4.

• Allen Institute for AI unveils Olmo 3 models that it says outperform open models like Stanford’s Marin and commercial open-weight models like Llama 3.1.

• Meta releases Segment Anything Model 3 for advanced image segmentation tasks.

• And Google DeepMind releases WeatherNext 2, a weather model that it says offers faster, more accurate two-week forecasts and includes more tools for energy traders.

Subscribe now

Elsewhere in the FAANG free-for-all:

• Google says the Gemini app is now able to detect images created or edited by Google AI, and that it plans to roll out verification of video and audio “soon”.

• Microsoft launches Agent 365, a framework that lets businesses deploy and manage AI agents like human employees, with dashboards showing telemetry and alerts.

• Microsoft integrates more AI features into Windows, including new Copilot skills, AI agents on the taskbar, writing assistance, and troubleshooting agents.

• Amazon launches a Video Recaps feature on Prime Video that uses AI to summarize a show’s key plot points with snippets and music, available in beta for US users.

• Yann LeCun says he is leaving Meta at the end of 2025 to build a new startup and continue his “Advanced Machine Intelligence research”, with Meta as a partner.

• And Meta Chief Revenue Officer John Hegeman is leaving after 17 years at Meta; Clara Shih, who joined in November 2024 to lead its Business AI unit, is also exiting.

Irrational Exuberance

Tech leaders are now openly questioning the trillion-dollar AI investment boom, with Google CEO Sundar Pichai warning of “elements of irrationality” in the current spending frenzy, the CEO of Klarna being “nervous about the size of [data center] investments,” and Peter Thiel’s hedge fund selling its entire $100 million Nvidia stake.

Between the lines:

• AI company valuations continue to soar - Nvidia’s $7T, Alphabet’s $3.5T, OpenAI’s $500B, Anthropic’s $350B, and xAI’s $230B - raising questions about whether AI-powered products will generate enough revenue to justify the massive capital outlays.

• Pichai’s candid admission that “no company is going to be immune” to a potential AI bubble burst - including Google - marks a notable shift from the unchecked optimism that has characterized tech leadership rhetoric.

• JPMorgan analysts estimate that $5 trillion in global AI infrastructure investment through 2030 would require generating an additional $650 billion annually - equivalent to every iPhone owner paying $35 more per month - to deliver reasonable returns to investors.

• Despite the warnings, tech giants are doubling down rather than pulling back - but supply chain constraints for critical components like transformers (booked through 2028) and power generation equipment are creating hard limits on how fast data centers can actually be built.

Elsewhere in AI infrastructure:

• The US DOE accelerates its approach to equipping national labs with AI supercomputers by working with Nvidia, AMD, and Oracle, which will pay some of the costs.

• AI server shipments from Taiwan to the US are set to double in 2025 compared to 2024, as Foxconn, Quanta, and Wistron ramp up production to meet demand.

• The US Commerce Department approves the sale of up to 35K Nvidia GB300 servers or their equivalents each to Abu Dhabi’s G42 and Saudi government-backed Humain.

• Elon Musk says xAI plans to develop a 500-megawatt data center in Saudi Arabia with state-backed AI startup Humain; the data center will rely on Nvidia chips.

• And Elon Musk, Jeff Bezos, and Sundar Pichai have each discussed building lunar and orbital AI data centers to leverage constant solar power and bypass regulations.

Elsewhere in OpenAI:

• OpenAI expands group chats in ChatGPT globally to all logged-in users on Free, Go, Plus, and Pro plans after piloting the feature in select regions.

• OpenAI announces ChatGPT for Teachers, designed for K-12 educators and school districts, which will be free to K-12 educators in the US through June 2027.

• OpenAI and Intuit sign a multiyear deal set to generate $100M+ in OpenAI revenue, in which Intuit will deepen its use of OpenAI models and launch ChatGPT apps.

• OpenAI says ChatGPT will now avoid em dashes if users tell it to, addressing the telltale sign that has signaled AI-written text.

• And OpenAI says GPT-5 has demonstrated the ability to accelerate scientific research workflows but can’t run projects or solve scientific problems autonomously.

States Rights

President Trump is reportedly preparing an Executive Order as soon as Friday that would give the federal government sweeping power to override state AI laws, including creating an “AI Litigation Task Force” under the attorney general to go after “woke” state AI regulation.

Why it matters:

• This would represent a dramatic assertion of federal power over AI regulation, with the Justice Department empowered to challenge state laws like California’s AI safety requirements and Colorado’s algorithmic discrimination protections.

• The “AI Litigation Task Force” would weaponize critical infrastructure funding - states could lose Broadband Equity Access and Deployment (BEAD) program money for rural broadband if they maintain AI regulations deemed “onerous.”

• But the EO wouldn’t pass without a fight - not only will there be lawsuits, but also intra-party dissent. High-profile Republicans - including Republican governors, senators, and MAGA supporters - are already openly opposing the President.

• Ultimately, a pro-AI stance is increasingly a risky political calculation for both parties: concerns about AI-related child suicides, job losses, and energy costs could become devastating attack lines in 2028.

Elsewhere in AI geopolitics:

• The European Commission’s postponement of a key part of the AI Act has become a case study for critics who say the EU prioritizes regulation over innovation.

• Key White House officials are pressing lawmakers to keep AI chip export restrictions to China out of the annual defense policy bill.

• Pro-AI super PAC Leading the Future is targeting New York State Assembly member Alex Bores, who co-sponsored the RAISE Act requiring AI safety protocols.

• An investigation reveals that as US Commerce Secretary Howard Lutnick pushes AI data centers and pressures US allies, his companies and sons are profiting from them.

• And the current AI strategies of China and the US are complementary, as China focuses on embodied AI and open source models rather than AGI like the US.

Elsewhere in AI anxiety:

• A study found that teen mental health chatbots like ChatGPT, Claude, Gemini, and Meta AI often failed to recognize signs of mental health conditions and gave general advice.

• AI is pressuring Poland’s economy, Europe’s top location for global business services, as companies lay off staff in Kraków’s BPO hub.

• Some experts question Anthropic’s claims of cyberattack breakthroughs using its tools, noting that white-hat hackers report modest gains from AI-aided hacking.

• Chinese toymaker FoloToy suspended sales of its GPT-4o-powered teddy bear after researchers found the toy gave kids harmful responses about fire and kinks.

• And Microsoft warns that Copilot Actions in Windows, now in beta and off by default, can infect devices and pilfer data, prompting concern from security researchers.

Things happen

Warner Music settles its lawsuit with AI music startup Udio. Soumith Chintala, PyTorch co-creator, joins Mira Murati’s lab. Poe launches group chats with up to 200 people across 200+ AI models. A look at Inception Point AI’s podcast network generating 3,000 podcasts per week. Cloudflare acquires Replicate, which hosts 50K+ AI models. Larry Summers resigns from OpenAI’s board after his Epstein emails surface. Elon Musk could “drink piss better than any human in history,” Grok says. Massive leak shows chatbot users turned yearbook pictures into AI porn. A researcher made an AI that completely breaks online surveys. DeepMind plans to open an AI research lab in Singapore. An overview of macro tech trends for 2026 as AI eats the world. At an AI conference, attendees said they’d short Perplexity over OpenAI. AI is a front for consolidation of resources and power. The AI bubble is bigger than you think. Claude launches structured outputs on its developer platform. Solving a million-step LLM task with zero errors. Heretic: Automatic censorship removal for language models. Google opens its biggest AI hardware hub outside the US in Taipei. Berkshire Hathaway discloses a $4.3B stake in Alphabet. Demis Hassabis pursued a Nobel as a DeepMind objective over revenue. Meta will assess employees on their “AI-driven impact” starting in 2026. Apple requires apps to disclose AI data sharing with third-party providers.

Last week’s roundup

Agentic espionage

Anthropic has disclosed what it believes is the first large-scale cyberattack executed almost entirely by AI agents, with a Chinese state-sponsored group using Claude Code to autonomously infiltrate roughly thirty global targets.

The big picture:

• The automation level marks a significant escalation: AI handled 80-90% of the operation with only 4-6 critical human decision points, performing reconnaissance, writing exploit code, harvesting credentials, and extracting data at thousands of requests per second.

• Weirdly, this may be a bullish signal for American AI labs. The attacks come as US export controls create acute semiconductor shortages in China, prompting companies to resort to smuggling and workarounds - implying that it may be more effective or reliable to use Anthropic’s models over Chinese ones.

• Ultimately, the incident underscores AI’s dual-use dilemma: the same tools designed to strengthen cybersecurity defenses can be weaponized to supercharge attacks, creating an arms race where defenders must maintain a “substantial permanent advantage” or risk losing ground to increasingly automated threats.

Subscribe now

Elsewhere in AI geopolitics:

• The UK proposes an amendment to the Crime and Policing Bill that would let “authorized testers“ proactively assess AI models for their ability to generate CSAM.

• A Munich court sides with Germany’s music rights society GEMA in a case against OpenAI, saying OpenAI can’t use song lyrics without a license.

• The European Commission plans to simplify some of its privacy rules, including GDPR, to boost AI growth and slash red tape for businesses in Europe.

• Anthropic open sources a method to score AI model political evenhandedness, with Gemini 2.5 Pro scoring 97%, Grok 4 at 96%, Claude Opus 4.1 at 95%, GPT-5 at 89%, and Llama 4 at 66%.

• OpenAI asks a US federal judge to reverse an order requiring it to turn over 20M anonymized ChatGPT chat logs in the NYT’s copyright lawsuit, citing privacy concerns.

• And an investigation traces how Shanghai-based AI startup INF Tech accessed advanced Nvidia chips in Jakarta, despite US export controls.

Elsewhere in AI anxiety:

• US schools are increasingly adopting AI-powered monitoring tools like GoGuardian to scan students’ chatbot conversations for signs of self-harm.

• A group of lawyers has documented 533 cases of AI misuse in legal filings, including fabricated case law citations, even as judges and bar associations permit AI use.

• Seven lawsuits in California claim ChatGPT encouraged dangerous discussions that led to suicides and harmful delusions.

• And in DeepSeek’s first public appearance since R1’s success, a senior researcher told a state-run conference he was pessimistic about AI’s impact on humanity.

End of an era

Yann LeCun, Meta’s chief AI scientist and Turing Award winner, is planning to leave the company to start his own venture.

The big picture:

• LeCun’s exit represents a fundamental clash of visions: he believes current large language models will never achieve human-level reasoning - yet Zuckerberg has bet Meta’s future on rapidly deploying LLM-based products.

• The Frenchman will be following another AI pioneer, Fei-Fei Li, in developing “world models” - AI systems that learn from spatial and video data rather than text. Li’s startup, World Labs, debuted Marble, its first publicly available model, which turns prompts and photos into editable 3D environments.

• With mounting Wall Street pressure on Zuckerberg’s $100B+ AI spending, Meta appears to be sacrificing long-term moonshots for short-term competitive positioning - disbanded the patient, academic approach of the FAIR lab in favor of exclusive teams poaching talent from OpenAI and Google.

Elsewhere in the FAANG free-for-all:

• LinkedIn rolls out AI-powered people search to Premium subscribers in the US, with plans to expand globally in the coming months.

• Alibaba plans to rename its AI app Tongyi to Qwen and add agentic AI shopping features for platforms like Taobao.

• A deep dive examines Microsoft’s AI strategy, including its OpenAI deal, data center investments, GitHub Copilot, MAI models, and Maia chip.

• Google unveils Private AI Compute, a cloud platform providing a secure space to run AI tools on devices, similar to Apple’s Private Cloud Compute.

• And Google rolls out Nano Banana image editing upgrades in Google Photos, including a “Help me edit” feature that lets users make edits using text or voice prompts.

Elsewhere in frontier models:

• OpenAI rolls out GPT-5.1 Instant, which is “warmer by default”, and GPT-5.1 Thinking, which is “easier to understand and faster”, starting with paid subscribers.

• Baidu unveils Ernie 5.0, an AI model to process and generate text, images, audio, and video, claiming it beats GPT-5-High and Gemini 2.5 Pro on some benchmarks.

• Google DeepMind unveils SIMA 2, a video-game-playing agent built on top of Gemini to navigate and solve problems inside 3D virtual worlds like Goat Simulator 3.

• OpenAI is piloting group chats in ChatGPT, with up to 20 users able to prompt ChatGPT in a shared space.

• And Meta releases Omnilingual Automatic Speech Recognition, a suite of AI models handling automatic speech recognition for 1,600+ languages, vs. OpenAI Whisper’s 99.

Moonshots vs Margins

Financial documents reveal that Anthropic expects to break even by 2028 while OpenAI won’t turn profitable until 2030, reflecting fundamentally different strategies in the AI race - one focused on sustainable growth, the other on ambitious moonshots.

The big picture:

• OpenAI is spending aggressively to maintain its lead, with projected losses of $74 billion in 2028 and commitments totaling $1.4 trillion over eight years. Anthropic is taking a more conservative path, focusing on enterprise customers (80% of revenue), avoiding costly ventures into image and video generation, and growing costs more in line with revenue.

• And yet - Anthropic is still getting into the infrastructure game, with a $50 billion plan to build AI data centers in partnership with Fluidstack, starting with facilities in Texas and New York.

• The divergence is testing investor patience at a critical moment, with tech stocks being punished over AI spending concerns and skepticism mounting over the industry’s massive financial commitments. Anthropic’s path to profitability may prove prescient - or OpenAI’s massive upfront investment could cement its market dominance if demand continues surging.

Elsewhere in AI infrastructure:

• Baidu unveiled two AI chips: the M100 for efficient MoE inference coming in early 2026, and the M300 for training super-large multimodal models coming in 2027.

• Several of Asia’s top tycoons and conglomerates are joining the data center race as tech giants plan $240B in APAC hyperscale expansion over the next five years.

• Microsoft will spend $10B to build a data center park in Sines, Portugal, in partnership with Portuguese developer Start Campus and UK startup Nscale.

• Meta says it will invest $600B in US infrastructure and jobs by 2028, with Zuckerberg telling Trump at a September dinner that Meta will spend “at least $600B.”

• And Elon Musk says Tesla is “probably going to have to build a gigantic chip fab“ for AI chips and “maybe, we’ll do something with Intel.”

Things happen

Intel CTO and AI officer Sachin Katti is leaving to join OpenAI. A inside look working at Cursor for 60 days. Six models traded $10K on crypto derivatives; Qwen3 Max lost $652 while GPT-5 lost $5,679. Amazon’s House of David Season 2 had 350 to 400 AI-generated shots. Cursor crosses $1B in annualized revenue. Free access to ChatGPT and Gemini suggests India may be the biggest bet for broad AI adoption. SoftBank sold its entire Nvidia stake for $5.8B in October. OpenAI is piloting group chats in ChatGPT in Japan, New Zealand, South Korea, and Taiwan. Pakistani newspaper mistakenly prints AI prompt with the article. AI is Dunning-Kruger as a service. ElevenLabs’s new Iconic Voice Marketplace for licensing AI versions of famous voices like Michael Caine’s. Kagi Search launches SlopStop, a community-driven AI slop detector. Nano Banana can be prompt engineered for nuanced AI image generation. How China ramped up its AI development from spring 2024 to catch the US. OpenAI’s Sora 2 floods social media with videos of women being strangled. Nasdaq had its worst 5-day run since April as AI stocks fell by about $800B.

Last week’s roundup

Agent provocateur

Amazon is suing Perplexity to block the startup’s Comet browser agent from making purchases on behalf of users on Amazon.com, marking one of the first major legal battles over AI agents’ ability to perform real-world tasks autonomously.

The big picture:

• This lawsuit could set critical precedents for how far “agentic AI” can go in acting on users’ behalf - Amazon argues Perplexity is committing computer fraud and violating its TOS by disguising AI agents as human users; Perplexity counters that agents should have “all the same rights and responsibilities” as the humans who deputize them.

• As AI shopping assistants proliferate, they may become a competitive threat to Amazon’s advertising empire. If AI agents shop autonomously for users, the lucrative business of selling prominent product placement in search results potentially loses its value.

• However, new research reveals that current models struggle with too many choices, are vulnerable to manipulation, and have poor collaboration skills even when unsupervised.

Subscribe now

Elsewhere in the FAANG free-for-all:

• Amazon is testing Kindle Translate, an AI tool that automatically translates books into other languages for self-publishing authors.

• Microsoft AI CEO Mustafa Suleyman says Microsoft plans to focus on superintelligence that prioritizes human control and will lead a new superintelligence team.

• Google says Ironwood, its seventh-generation TPU, will launch in the coming weeks and is more than 4x faster than its sixth-gen TPU.

• Apple is finalizing a deal to pay Google ~$1B per year for a 1.2T-parameter Gemini model to help power the new Siri, which is on track for next spring.

• And Google unveils Project Suncatcher, aiming to launch two solar-powered satellites carrying four TPUs each into low Earth orbit in 2027 with partner Planet Labs.

Backstop backlash

OpenAI CFO Sarah Friar sparked a swift political backlash after discussing the idea of a federal “backstop” to help finance the buildout of AI compute.

Between the lines:

• The remarks prompted White House AI czar David Sacks to declare there will be “no federal bailout for AI,” signaling that the Trump administration is willing to help through regulatory relief rather than financial support.

• But Friar’s word choice - and hasty retreat - reveals the precarious position OpenAI finds itself in after signing massive infrastructure deals that have raised serious questions about its ability to pay for them. Even CEO Sam Altman stepped in to try to defuse the situation.

• How we resolve this tension remains an open question - the government wants American companies to dominate globally, but there’s clearly political liability in creating AI giants that are “too big to fail.”

Elsewhere in OpenAI:

• OpenAI and AWS signed a seven-year, $38B deal for AI compute, including training models using Amazon’s data centers and CPUs.

• OpenAI released a blueprint for teen AI safety standards with five suggestions the company hopes lawmakers will use in crafting regulations.

• Sam Altman discussed managing OpenAI’s growth, delegating work, hiring hardware talent, GPT-6 enabling scientific research, and AI’s societal challenges in a Q&A.

• Court documents reveal that in a deposition, Ilya Sutskever discussed conflicts at OpenAI that he sent to board members before Sam Altman’s firing, his OpenAI exit, and more.

• And CODA, a Japanese anti-piracy group representing Studio Ghibli and Square Enix, demands that OpenAI stop using their copyrighted content to train Sora 2.

Elsewhere in AI infrastructure:

• Meta, Alphabet, Microsoft, and Amazon announced plans to increase their combined AI spending to $400B in 2026 in their latest earnings reports.

• Tech companies, including Meta and xAI, are using SPVs to raise tens of billions for AI data centers, allowing them to keep the debt off their balance sheets.

• Microsoft CEO Satya Nadella says the company faces a power shortage, not a compute one, which could leave chips sitting in inventory that can’t be plugged in.

• Oxford Economics reports that the AI boom is lifting global trade, with approximately 60% of US data center capex spent on imported IT gear mainly from Taiwan, South Korea, and Vietnam.

• Samsung is partnering with Nvidia to build an “AI Megafactory” and deploy over 50,000 of Nvidia’s most advanced GPUs to embed AI in its chipmaking process.

• Google plans to build an AI data center on Australia’s remote Christmas Island after signing a cloud deal with the Department of Defense earlier in 2025.

Huawei or the highway

Despite reaching a market cap of $5 trillion, Nvidia’s future now looks less rosy, as the Trump administration officials declined to sell advanced Blackwell chips to China ahead of the Trump-Xi summit in South Korea.

Where we’re headed:

• The decision comes as China bans foreign AI chips from state-funded data centers - potentially affecting over $100 billion in projects - while simultaneously slashing energy costs by up to 50% for facilities using domestic semiconductors (Huawei and other Chinese semiconductors require 30-50% more electricity than Nvidia’s chips to perform the same tasks).

• Huang’s unusually blunt warnings - criticizing Western “cynicism” and predicting China’s victory in the AI race - reflect Nvidia’s dire position in a market where its share has plummeted from 95% in 2022 to zero today.

• With Trump planning an April visit to China and Huang vowing to continue his late-night lobbying calls, the current stalemate appears temporary - but each side’s hardening position suggests the AI cold war is entering a new phase of permanent technological decoupling rather than managed competition.

Elsewhere in AI geopolitics:

• Senators Mark Warner and Josh Hawley unveiled a bipartisan bill that would require companies and US agencies to submit quarterly reports on AI-related job impacts.

• At APEC, Xi Jinping proposed a World Artificial Intelligence Cooperation Organization for global AI regulation, which Chinese officials said could be based in Shanghai.

• Google removed its AI model Gemma from AI Studio after Senator Blackburn said it fabricated sexual misconduct accusations against her.

• The US CBP signed a two-year deal with AI logistics company Altana to use its platform for real-time trade enforcement and forced labor detection.

• And Nvidia will supply 260K+ of its most advanced AI chips to the South Korean government and major companies, including Samsung, Naver, and SK Group.

Elsewhere in AI anxiety:

• xAI and Gab position their ideology-driven AI chatbots Grok and Arya as alternatives to mainstream chatbots accused of liberal bias like ChatGPT and Gemini.

• Some podcasters are embracing AI voice clones from ElevenLabs and other startups to augment, translate episodes, and even replace in-studio performances.

• The UK’s High Court ruled against Getty in its lawsuit against Stability AI over its image generator, with the judge calling the ruling “extremely limited in scope.”

• Coca-Cola released new AI-generated ads for the 2025 holidays after finding people didn’t know or care about the use of AI, despite criticism for its 2024 ads.

• Online harassers are using AI tools to create more realistic death threats, posting AI-generated images and sounds that simulate victims in violent situations.

• And arXiv will step up its moderation of CS review papers not vetted by an academic journal or conference after a surge in AI-generated papers.

Things happen

Perplexity launches a free AI-powered patent research tool in beta. Google in talks to value Anthropic at $350B+. Small AI models do the real work. Grokipedia mirrors Wikipedia except on politics, but AI makes it better on obscure entries. Chinese startup Moonshot releases Kimi K2 Thinking, claiming it beats GPT-5 for $4.6M. OpenAI now has 1M+ business customers globally. Lessons on deploying AI agents in production. Gemini Deep Research can now draw on your Gmail, Drive, and Chat to generate reports. Nebius unveils Token Factory to compete with AWS and Azure. AI designs functional antibodies from scratch, no animals required. OpenAI’s o1 analyzes languages as well as a human expert. Deutsche Bank is exploring strategies to hedge its exposure to data centers. What jobs AI is replacing today. AI coding ability doubles every six months, but real-world tasks have a messiness tax. You should write an agent. How AI turned Google Cloud from also-ran to growth driver. The man who invented AGI. Getty and Perplexity sign a multi-year licensing agreement for image distribution. A profile of Common Crawl, the nonprofit scraping the web for AI training data. The future of advertising is AI-generated ads personalized to you. AI’s dial-up era. How I use every Claude Code feature.

Last week’s roundup

Layoff season

Even as core economic data suggests that AI is not having as big of an impact on the economy as previously predicted, it’s hard to ignore the headlines of Fortune 500 companies shedding tens of thousands of jobs.

The big picture:

• Untangling the root cause here is messy: even if AI isn’t replacing those workers today, that doesn’t stop CEOs from doing layoffs in the hope that AI can retroactively replace them tomorrow (though we’re already seeing some cautionary tales here).

• It’s also likely that many of these layoffs are the result of some combination of pandemic-era overhiring and an increasingly complex macroeconomic environment - though the political climate may make it challenging to say the latter part publicly.

• But it is clear that we’re seeing companies attempt to adopt AI usage (if not outright AI automation) at a far faster speed than previous technology shifts - leaving less time for workers to retrain and labor markets to adjust, and potentially creating acute economic and social disruption.

Subscribe now

Elsewhere in the FAANG free-for-all:

• Amazon opens Project Rainier, an $11B AI data center on 1,200 acres in Indiana that trains and runs Anthropic’s AI models using 500K+ Amazon Trainium 2 chips.

• GitHub plans to launch Agent HQ, a hub to let developers access third-party AI coding agents, including Claude and OpenAI’s Codex, in the coming months.

• Adobe unveils Project Moonlight, an AI agent on its Firefly platform designed to act as a creative director for social media campaigns via text prompts.

• And Grokipedia, a Grok-based encyclopedia, launches with 885K+ articles, drawing from Wikipedia content; Elon Musk called it “a massive improvement” over Wikipedia.

Elsewhere in AI anxiety:

• Scale AI and CAIS’ Remote Labor Index finds the best AI performed less than 3% of tasks when measuring AI models’ ability to automate freelance work.

• Employees are increasingly using AI to create ultra-realistic fake expense receipts, with AppZen and others attributing the rise to advances in image generation models.

• AI startups are aiming to address loneliness among the elderly, with a study predicting the AI market for aging and elderly care will grow to over $43B in 2025.

• An a16z-backed startup is selling thousands of ‘synthetic influencers’ to manipulate social media as a service.

• An ex-OpenAI staffer of four years says it isn’t doing enough to protect users, especially those with mental health issues, and needs to offer more than words.

OpenAI is for profits now

OpenAI completed a major restructuring that has its nonprofit arm owning 26% (though still retaining control) of the for-profit company. The deal also restructured Microsoft’s ownership agreement, with the tech giant retaining 27% ownership, valued at $135 billion.

Between the lines:

• The agreement reveals both companies hedging their AGI bets: Microsoft can now pursue AGI independently or with other partners, while OpenAI gains freedom to release open-weight models and shop around for compute infrastructure - though Microsoft does retain IP rights through 2032, and OpenAI has committed to $250B in Azure spend.

• The political dimension proved crucial: OpenAI appears to have gotten regulatory approval for the deal by promising to remain in California and emphasizing how AI is currently “the economic heart” of the state’s economy.

• Ultimately, the deal is preparing for a future where OpenAI operates as a standalone public company with multiple partners rather than as Microsoft’s captive AI lab - as reports suggest OpenAI is preparing for an IPO as early as late 2026, valuing the company at up to $1 trillion.

Elsewhere in OpenAI:

• The Sora app adds new features, including “character cameos” to create AI deepfakes of pets and illustrations, as it begins selling usage credits (free users are limited to 30 daily generations).

• OpenAI launches Aardvark, a GPT-5-powered autonomous cybersecurity research agent that can identify and help patch vulnerabilities, in private beta.

• A new “company knowledge” ChatGPT update lets Business, Enterprise, and Education users search for data in connected apps like Slack and GitHub.

• OpenAI is developing AI tools to generate music from text and audio prompts, including capabilities such as adding guitar accompaniment to vocal tracks.

• And PayPal signs a deal with OpenAI to let PayPal buyers and sellers complete transactions through ChatGPT starting in 2026, using the Agentic Commerce Protocol.

Elsewhere in AI geopolitics:

• OpenAI urges the US to commit to building 100GW of new energy capacity each year to stay ahead of China in AI, calling “electrons are the new oil.”

• President Trump says he and Xi Jinping discussed Nvidia and other chipmakers’ access to China, but didn’t discuss approving sales of Blackwell chips.

• The US AI data center boom is consuming capital, power, and people that the rest of the economy needs, crowding out Trump’s manufacturing revival plans.

• Papers, patents, and tenders show China’s military is integrating DeepSeek and Qwen models in weapons like AI-powered drones and continues to use Nvidia chips.

• And Saudi Arabia is pouring money into data centers and working with US and Chinese tech giants, landing its AI ambitions in the middle of a geopolitical tussle.

The five trillion dollar company

On the heels of its GTC in DC event, Nvidia has become the world’s first company to reach a $5 trillion market capitalization, cementing its position as the most valuable corporation in history.

Why it matters:

• The concentration of such enormous market value in a single chipmaker has raised questions about market stability and monopolistic dynamics in the AI supply chain, particularly as Nvidia’s fortunes remain tied to sustained enterprise spending on AI systems whose long-term returns remain unproven.

• Of course, Nvidia’s CEO pushed back against AI bubble concerns, announcing partnerships across industries and projecting half a trillion dollars in revenue from its latest chips - spanning humanoid robots (Foxconn), cybersecurity (CrowdStrike), defense/logistics (Palantir), and pharma (Eli Lilly), aiming to broaden its revenue base as competition intensifies.

• And the conference’s patriotic messaging and Trump references reflect Nvidia’s delicate geopolitical balancing act: championing American manufacturing while pursuing “sovereign AI” deals in Europe and seeking White House help to restore access to the Chinese market that’s cost them billions in lost revenue.

Elsewhere in Nvidia:

• Nvidia invested in 59 AI startups in 2025 so far, including 10 in the past two months, surpassing 55 investments in 2024 and 12 investments in 2022.

• Jensen Huang says the US needs “finesse” to maintain its AI lead over China, and Nvidia is in an “awkward place” as President Trump meets Xi Jinping later this week.

• The company unveiled BlueField-4 data processing unit, combining a 64-core Grace CPU and ConnectX-9 SuperNIC for 800 Gbps of network throughput, launching in 2026.

• Nvidia’s Blackwell GPUs are now in full production in Arizona, after being solely manufactured in Taiwan previously.

• And Nvidia announced NVQLink, a high-speed interconnect that is the “Rosetta Stone connecting quantum and classical supercomputers.”

Elsewhere in AI infrastructure:

• Google, Meta, and Microsoft collectively spent nearly $80B on AI infrastructure in Q3, dividing the market on whether they can translate huge capex into income.

• The AI buildout, surpassing railroads and the internet, moves Big Tech from being asset-light to asset-heavy, which has historically produced inferior returns.

• Sam Altman says OpenAI has committed to spending about $1.4T on infrastructure so far, equating to roughly 30GW of data center capacity.

• An analysis of 30+ US data center proposals in 14 states found that in most cases, local officials signed NDAs and worked with shell companies to hide details.

• And Indonesia and other developing countries are becoming AI hubs amid the multitrillion-dollar spending boom, in part driven by an “AI decolonization” philosophy.

Things happen

OpenAI releases gpt-oss-safeguard, its open-weight reasoning models for safety classification tasks. Canva launches a foundational design model that generates editable layered designs. YouTube is using AI to upscale videos to “HD.” Grammarly rebrands as Superhuman. AI can code, but it can’t build software. UMG announces partnerships with Stability AI and Udio to create fully licensed AI-generated music. Qualcomm unveils two AI inference chips, the AI200 and AI250. The AI boom will spark a “wildfire” that wipes out many companines. I’m drowning in AI features I never asked for and I hate it. Using AI to negotiate a $195k hospital bill down to $33k. Anthropic expands Claude for Financial Services with Excel integration. IBM releases open-source Granite 4.0 Nano AI models. JPMC staff can officially use an LLM for year-end performance reviews. Andrew Cuomo uses AI MPREG Schoolhouse Rock to attack Mamdani. Character.ai limits under-18s to two hours of chat ahead of November ban. Cal State to become “AI-empowered” university via partnerships with Amazon and ChatGPT. China’s MiniMax releases MiniMax M2, an open-source model optimized for AI agents and coding. Extropic shares its first chip using probabilistic bits that it says can be 10,000x more energy efficient. It’s insulting to read AI-generated blog posts. The new calculus of AI-based coding.

Last week’s roundup

The browser wars continue

OpenAI launched ChatGPT Atlas, the latest AI-powered web browser with “agent mode” that can autonomously navigate websites and complete multi-step tasks on users’ behalf.

The big picture:

• Atlas builds on OpenAI’s previous agentic AI attempts like Operator and ChatGPT Agent, featuring personalized memory, a persistent ChatGPT companion in split-screen mode, and “cursor chat” for in-line text editing. But early testing reveals a slow, clunky experience, with significant limitations on session length.

• The security implications are substantial, as even OpenAI’s CISO acknowledged: prompt injection attacks remain an “unsolved frontier security problem” that could allow malicious websites to hijack the agent and leak credentials.

• By launching a browser, OpenAI is directly competing with Google, Microsoft, and Apple on their home turf - not to mention Perplexity’s Comet and Opera’s Neon. However, whether users want an AI intermediary for everyday web tasks - and whether they’ll trust it with sensitive actions - remains an open question.

Subscribe now

Elsewhere in OpenAI:

• OpenAI acquires Software Applications Incorporated, an AI-powered user interface startup for macOS that raised $6.5M from Sam Altman and others.

• The company has hired more than 100 ex-investment bankers, paid $150 per hour, to train its AI to build financial models as part of its secretive Project Mercury.

• Sam Altman’s deal spree to secure vast compute for OpenAI pits Silicon Valley giants against each other as they race to cash in, making OpenAI too big to fail.

• OpenAI says it will work with SAG-AFTRA, CAA, UTA, and others to crack down on Sora 2 deepfakes, following concerns from Hollywood.

• And Microsoft, OpenAI, and Anthropic are providing millions to the American Federation of Teachers to build AI training hubs aimed at educating 400,000 teachers.

Shrinking Superintelligence

Meta is cutting roughly 600 jobs from its AI Superintelligence Lab (out of several thousand), as Mark Zuckerberg works to streamline AI operations that have become “overly bureaucratic.”

Between the lines:

• This isn’t a cost-cutting measure - Meta is betting that starting fresh with new leadership and a smaller, “more load-bearing” team will outperform years of accumulated institutional knowledge, including established teams from FAIR, product AI, and infrastructure.

• The restructuring comes amid reports that top AI researchers across Silicon Valley are working 80-100 hour workweeks in a wartime-like effort to compress decades of scientific progress into years - turning the small cluster of elite AI researchers into millionaires who have no time to spend their fortunes.

• Meta’s reorg reveals anxiety about falling behind, especially considering the layoffs included over 100 employees in its risk review organization - the team responsible for ensuring compliance with FTC privacy requirements.

Elsewhere in the FAANG free-for-all:

• Anthropic and Google announce their cloud partnership worth tens of billions of dollars, giving Anthropic access to 1M TPUs and 1GW of capacity in 2026.

• Microsoft unveils Mico, a new character for Copilot’s voice mode that responds with real-time expressions, and Copilot Groups, enabling up to 32 people to collaborate in a session.

• Amazon is testing AR glasses for delivery drivers, using AI and computer vision to help them scan packages, follow walking directions, and get proof of delivery.

• YouTube launches its likeness detection tech, letting eligible creators in its Partner Program request the removal of AI-generated content with their likeness.

• Meta plans to add new Instagram teen safety tools in 2026 that let parents block teens from chatting with AI characters and get “insights” from teens’ chats.

• And WhatsApp updates its Business API terms to ban general-purpose chatbots starting January 15, 2026, affecting WhatsApp assistants of OpenAI, Perplexity, and others.

Elsewhere in AI anxiety:

• A teen’s parents allege OpenAI loosened ChatGPT’s suicide-talk rules to boost engagement before their son died by suicide using a method discussed with ChatGPT.

• Reddit filed a lawsuit against Perplexity and three data scraping companies, accusing them of illegally stealing its data by scraping Google search results.

• Researchers detailed systemic vulnerabilities in AI agentic browsers, including Perplexity’s Comet and Fellou, related to indirect prompt injection attacks.

• An EBU/BBC study found that 45% of responses from top AI assistants misrepresented news content with at least one significant issue, and 31% showed serious sourcing problems.

• And AI-generated ‘poverty porn’ fake images are being used in social media campaigns from leading health NGOs.

DeepSeek diplomacy

While American AI companies compete for lucrative Western markets, DeepSeek and Huawei have been expanding into Africa with affordable, open-source AI models to build long-term influence across the continent.

Why it matters:

• When it comes to African businesses, Chinese open-source models are available at a fraction of the price of Western competitors: DeepSeek charges $1.10 per million output tokens compared to OpenAI’s $15.

• This mirrors China’s Belt and Road Initiative: it’s not about immediate profit, but about securing future customers, soft power, and vast data troves while African nations lack alternatives.

• And it’s not just APIs: ByteDance’s chatbot Cici is quietly gaining traction in the UK, Mexico, and Southeast Asia. But US companies won’t give up emerging markets without a fight: ChatGPT Go, OpenAI’s sub-$5 subscription for India and Southeast Asia, is also coming to Africa.

Elsewhere in frontier models:

• DeepSeek released DeepSeek-OCR, a vision language model designed for efficient vision-text compression that enables longer contexts with less compute.

• Anthropic announced Claude Life Sciences, a new feature that integrates Claude AI models with lab tools like Benchling to boost research efficiency.

• Google plans to release Gemini 3 in December, which is expected to be more performant across the board, especially in coding and multimodal generation.

• Anthropic’s Claude memory feature is rolling out to Pro and Max subscribers after initially being available for Team and Enterprise users.

• And in an experiment, GPT-4o, Claude Sonnet 4.5, and DeepSeek-V3.2-Exp expressed secular, Western liberal values regardless of the language of the questions.

Elsewhere in AI geopolitics:

• Dario Amodei addresses “inaccurate claims” about Anthropic’s policy stances after David Sacks said the “real issue” is “Anthropic’s agenda to backdoor Woke AI.”

• Prince Harry and 800+ public figures sign a Future of Life Institute statement urging a ban on AI superintelligence development until it can be deployed safely.

• A look at the social debates in Chile over AI investments, emblematic of clashes happening globally over balancing economic growth with environmental concerns.

• Senate Republicans share an AI-generated video falsely depicting Chuck Schumer saying on camera comments he made in a print interview on the government shutdown.

• And US military has adopted an aggressive push to embrace AI, while the top US Army commander in South Korea says “Chat and I” have become “really close lately.”

Things happen

The AI boom’s electricity demand has caused a gas turbine supply crunch. EA partners with Stability AI for “smarter paintbrushes”. Krafton unveils an “AI First” strategy. OpenAI diversifies chip suppliers with Broadcom deal. Kayak launches AI Mode. Japanese stores use robots piloted remotely by Filipino workers to train AI models. Andrej Karpathy: AGI is a decade away. OpenAI and talent agencies discuss Sora. Microsoft won’t provide erotica AI services. GM to roll out Gemini in vehicles starting in 2026. The AI boom pushes SF rents up the most in the US. Replacement.ai. Pocketpair won’t fund games that use generative AI. Alibaba cut Nvidia GPU use by 82%. AI has a cargo cult problem. Dutch watchdog: don’t use AI to tell you how to vote. LLMs can get “brain rot”.

Last week’s roundup

ChatGPT After Dark

Sam Altman discussed coming updates to ChatGPT, including a friendlier personality, as well as “erotica for verified adults.”

The big picture:

• People are really, really focused on the erotica thing. And on the one hand, I get it - clearly “AI erotica” falls far short of building AGI, and it’s an easy dunk for cynics.

• But consider the alternative: as personal or sensitive usage of AI becomes more common, it’s unclear that companies like OpenAI should be the moral arbiters of what adults are allowed to chat about.

• I see parallels to the culture wars that continue to rage over terms like “free speech” and “censorship” on social media, which is a reminder that almost no big tech company has this figured out.

• That said, there is a fine line here (and I don’t mean to suggest otherwise) - parasocial relationships and emotional attachments are a real concern with modern chatbots, and we ought to have shared norms (if not laws) around how far these products can be optimized and marketed.

Elsewhere in AI anxiety:

• South Korea rolled back its AI textbook program after just four months following complaints about inaccuracies and extra workload.

• AI is replacing workers in India’s business process management sector, which employs 1.65M people, with conversational AI startup LimeChat saying it automated 5,000 jobs.

• Sora videos depicting dead celebrities sparked backlash from families, with OpenAI saying reps of recently deceased public figures can request likenesses be blocked.

• Judges in Latin America are struggling to rule on cases of AI-generated images even as courts increasingly adopt AI tools to clear case backlogs.

• The G20’s Financial Stability Board says regulators are in the early stages of tracking risks posed to the financial system by AI’s rapid adoption.

• And Wikipedia says AI is causing a dangerous decline in human visitors to the site.

Subscribe now

Elsewhere in OpenAI:

• OpenAI has proposed a “sign in with ChatGPT” feature to some companies, which could transfer the cost of using OpenAI’s models to their customers.

• Black hole theoretical physicist Alex Lupsasca is the first person to join the OpenAI for Science initiative led by Kevin Weil.

• OpenAI is making a five-year business plan to meet $1T+ spending pledges, currently booking ~$13B in ARR with 70% from consumer ChatGPT use.

• A new Expert Council on Well-Being and AI with eight experts will help guide OpenAI’s work on ChatGPT and Sora.

• And OpenAI and Broadcom agree to co-develop and deploy 10GW of custom AI chips to run OpenAI’s models over four years in a deal worth multiple billions.

Claude Haiku 4.5

Anthropic has released Claude Haiku 4.5, its newest lightweight model that matches the performance of Sonnet 4 while operating at one-third the cost and twice the speed.

Between the lines:

• What was cutting-edge AI capability in late 2024 is now available as a budget option in early 2025, illustrating the breakneck pace at which AI capabilities are becoming commoditized and accessible - this marks Anthropic’s third major launch in just two months (following Sonnet 4.5 and Opus 4.1).

• Anthropic is positioning Haiku as part of an “agent toolbox” where multiple models work together: sophisticated models like Sonnet handle complex planning while lightweight Haiku agents execute tasks in parallel.

• The release comes at the same time as Skills for Claude, a new usage pattern with folders of instructions, scripts, and resources that Claude can load. Though simple, the Skills setup likely unlocks additional agentic capabilities for Claude Code.

Elsewhere in frontier models:

• Google introduced Veo 3.1, with improved audio output and stronger prompt adherence, and rolls out new updates to its AI video editor Flow.

• Microsoft unveiled MAI-Image-1, its first text-to-image AI model developed in-house, and says it “excels” at photorealistic imagery, like lighting and landscapes.

• Sora 2 users can now generate videos up to 15 seconds in the app and on the web, while Pro users can generate videos up to 25 seconds on the web.

• And OpenAI says GPT-5 instant and GPT-5 thinking reduce political bias by 30% compared to its prior models, and show “greater robustness to charged prompts”.

Elsewhere in the FAANG free-for-all:

• Apple executive Ke Yang, who was recently appointed to lead the AKI team developing AI-driven web search for Siri, is leaving for Meta.

• Google updated NotebookLM’s Video Overviews with six new Nano Banana-powered visual styles and a “Brief” format to help users quickly grasp core ideas.

• Meta announced a partnership with Arm to power AI ranking and recommendation systems across its family of apps using Arm-based data center platforms.

• xAI is building world models for use in gaming and robotics, and has hired two AI researchers from Nvidia to work on them.

• And Thinking Machines Lab co-founder Andrew Tulloch joins Meta, which has called an August report that he was offered a $1.5B+ package by Zuckerberg “ridiculous”.

The data center backlash

As AI infrastructure investment accelerates, small towns across America are increasingly rejecting data center proposals - and in at least one case, becoming the target of lawsuits.

What’s new:

• Saline Township, Michigan, initially rejected a 250-acre data center proposal, only to face a lawsuit from the developer claiming exclusionary zoning. The town ultimately settled, unable to sustain a prolonged legal battle despite overwhelming resident opposition.

• Other communities are taking more aggressive stances: St. Charles, Missouri enacted a year-long moratorium after a secretive 440-acre proposal near drinking water wells, while Lordstown, Ohio is proposing an outright permanent ban despite welcoming a Stargate-affiliated manufacturing facility.

• The backlash marks a significant shift in public awareness and sentiment about data centers’ downsides - including water and electricity consumption, noise pollution, lack of long-term job creation, and rising energy costs for residents - even as state governments eagerly court AI investments.

Elsewhere in AI data centers:

• Meta is investing $1.5B+ to set up a 1GW data center in El Paso, Texa and is working on an almost $30B financing package for its Louisiana data center.

• Google will invest $9B in South Carolina through 2027 to scale up data centers in Berkeley County and Dorchester County, with another $15B going to India to set up a 1GW data center in the southern state of Andhra Pradesh.

• Nvidia partnered with startup Firmus on Project Southgate, a ~$2.9B undertaking to build 150 MW of renewable energy-powered AI data centers across Australia.

• An investment consortium including BlackRock, Nvidia, xAI, and Microsoft will acquire Texas-based Aligned Data Centers from Macquarie in a $40B deal.

• UK-based cloud provider Nscale signed an up to $14B deal with Microsoft to deploy ~104K Nvidia GB300 chips in Texas within 18 months and 12,600 GPUs in Portugal.

• And CoreWeave and AI coding startup Poolside plan a 500-acre, natural gas-powered data center on a Texas ranch, with Poolside raising $2B at a $14B valuation.

Elsewhere in AI geopolitics:

• California Governor Gavin Newsom signed AB 1043, which requires device makers and app stores to verify users’ ages, and SB 243, mandating safety protocols for AI chatbot companions.

• The Senate passed a measure requiring Nvidia and AMD to prioritize US customers over China for advanced AI chip sales as part of its annual defense policy bill.

• China tightened customs checks on chip imports, starting with Nvidia’s H20 and RTX Pro 6000D, after urging local companies not to buy Nvidia’s products.

• And White House AI czar David Sacks accused Anthropic of running a “regulatory capture strategy“ based on fear-mongering in response to co-founder Jack Clark’s AI policy essay.

Things happen

Anthropic is on track to hit $9B in annual revenue by end of 2025. ByteDance’s Doubao became China’s most popular AI app in August. Andrej Karpathy unveils nanochat, a full-stack LLM deployable in 4 hours. Anduril’s AI-powered helmet will replace Microsoft’s IVAS for the Army. Salesforce unveils Agentforce 360 and positions Slack as an “agentic OS.” Slack is testing an AI-based Slackbot, rolling out end of 2025. Young people in China turn to AI chatbots for therapy. Cobots are bringing automation to small US factories. Indonesia’s film industry embraces AI tools for Hollywood-style movies. Figure AI’s Figure 03 robot, its first mass-producible humanoid. SemiAnalysis launches InferenceMAX, an open-source LLM benchmark. Reddit’s AI suggests users try heroin. Lawyer caught using AI while explaining to court why he used AI. Man stores AI-generated robot porn on government computer, loses nuclear secrets access. What happened when AI came for craft beer. Commonwealth Fusion Systems partners with DeepMind on Torax plasma simulator. Pinterest adds tools to limit AI-generated content in feeds. Microsoft only lets you opt out of AI photo scanning 3x a year. Beliefs true for software but false for AI. After the AI boom: what might we be left with? How much revenue justifies the current AI spend? State of AI Report 2025. LLMs are getting better at individual characters.

Last week’s roundup

OpenAI: DevDay

OpenAI’s annual developer conference, DevDay, was this week.

The announcements:

• ChatGPT Apps SDK, built on MCP, to let developers create apps inside of ChatGPT and launched with Canva, Coursera, Figma, Spotify, Zillow, and more.

• AgentKit, a toolkit for building and deploying AI agents, including a no-code Agent Builder interface.

• New Codex features: a Slack integration for delegating tasks or asking questions, a Codex SDK, and new admin tools

• And API updates, including GPT-5 Pro, Sora 2 (in preview), and gpt-realtime-mini, a voice model that is 70% cheaper than gpt-realtime.

The big picture:

• One easy comparison here is Facebook: turn a popular product into a “front door to the web” by letting developers build on top of it, then monetize through revenue-sharing or promoted placements. However, the technology arguably has the potential for more harm, given the sensitive nature of many ChatGPT conversations.

• Another is Microsoft: OpenAI’s strategy of doing everything - consumer, enterprise API, infrastructure deals - positions it as the linchpin of the entire AI bubble and mirrors Windows’ peak when Microsoft could answer every strategic question with “Yes,” before mobile forced focus.

• But perhaps Sam Altman’s preferred comparison is Apple: he took to the stage with Jony Ive to discuss their collaboration on a “family of devices” - though specifics remain scarce beyond a late 2026 potential launch and a philosophy focused on human wellbeing over productivity.

Subscribe now

Elsewhere in the FAANG free-for-all:

• Adobe launches AI agents tailored for B2B marketers within its Adobe Experience Platform, after launching AI agents aimed at consumer marketing in September.

• AWS launches Quick Suite, a chatbot and set of AI agents that can analyze sales data, produce reports, and summarize web content, set to replace Q Business.

• Google Cloud launches Gemini Enterprise, designed to help employees automate tasks and generate content across departments, priced at $30 per user per month.

• Google debuts an AI bug bounty program offering security researchers up to $30K for finding prompt injections, jailbreaks, and alignment issues in its products.

• Google DeepMind unveils CodeMender, an AI agent that detects, patches, and rewrites vulnerable code to prevent exploits by leveraging Gemini Deep Think models.

• And Anthropic and IBM partner to make Anthropic’s Claude models available in IBM’s latest IDE for large businesses, with IBM aiming to add Claude to more products soon.

OpenAI: Sora

The Internet is continuing to deal with the release of Sora, OpenAI’s AI-powered TikTok competitor. As Sora-created content continues to go viral, the company is releasing new guardrails and attempting to preempt potential legal challenges.

Between the lines:

• Initially, the company took a “ask for forgiveness, not permission“ approach on copyright, requiring creators to opt out rather than opt in to using their likenesses. And indeed, users were generating controversial content featuring copyrighted characters and public figures.

• But in the days since launch, OpenAI has backpedaled: first by adding granular controls for rightsholders over how characters can be used, and then by rapidly expanding content restrictions (now, even public domain characters like Winnie the Pooh and Dracula are blocked).

• Part of the challenge seems to be Sora’s rapid growth - despite being invite-only, the app hit the top of the App Store this week. It has reportedly reached 1 million downloads in under five days, faster even than ChatGPT.

• And there’s also the matter of AI-powered misinformation: even though Sora 2 includes visual watermarks on all generated videos, numerous websites have emerged that can remove these watermarks in seconds.

Elsewhere in frontier models:

• Google releases the Gemini 2.5 Computer Use model, built on Gemini 2.5 Pro’s capabilities to power agents that can interact with UIs, in preview via the API.

• Samsung introduces the Tiny Recursion Model, a 7M-parameter model that can outperform LLMs 10,000x larger, like Gemini 2.5 Pro and o3-mini, on specific problems.

• Huawei’s Zurich Lab unveils SINQ, an open-source quantization method that it claims can reduce LLM memory use by 60-70% without significant quality loss.

• And OpenAI updates GPT-5 Instant to better recognize and support people in distress, with ChatGPT routing such sensitive parts of conversations to the model.

OpenAI: AMD

OpenAI struck a major deal with AMD to deploy 6 gigawatts of AI chips over multiple years, and potentially giving the AI company a 10% stake in the chipmaker. But the deal is only adding to fears of a potential AI bubble.

Why it matters:

• The deal structure is novel - OpenAI can purchase up to 160 million AMD shares with tranches vesting as capacity milestones are met, culminating when AMD stock hits $600.

• OpenAI is aggressively diversifying its supply chain - including a $100B Nvidia deal, $10B Broadcom partnership, and deals with Samsung and SK Hynix - and is also striking multi-billion-dollar deals on compute with companies like Oracle ($300B) and CoreWeave ($22B).

• These enormous capital commitments raise critical questions about AI economics: whether these investments can generate proportional returns, and what happens if the current wave of AI enthusiasm wanes before these trillion-dollar bets pay off - especially as more and more deals are of a circular nature.

• It doesn’t help that Big Tech companies are increasingly financing their AI infrastructure buildouts through private debt and special purpose vehicles that keep borrowing off their balance sheets - a tactic reminiscent of the subprime mortgage crisis.

Elsewhere in AI anxiety:

• A study finds that as few as 250 malicious documents can produce a “backdoor” vulnerability in an LLM, regardless of model size or training data volume.

• A survey of 1,000 US highschool students finds that nearly one in five say they or someone they know has had a romantic relationship with AI.

• MrBeast, who is set to earn an estimated $85M in 2025, says AI-generated videos could threaten creators’ livelihoods, calling it “scary times” for the industry.

• A study found that by late 2024, around 24% of English-language corporate press releases and 14% of UN press releases used LLM-assisted writing.

• Deloitte says it will partially refund payment for an AU$439K Australian government report with multiple errors, after admitting it was partly produced by AI.

• And an investigation finds rampant caste bias in ChatGPT and Sora, with a researcher also finding caste bias in Sarvam AI, which touts itself as a sovereign AI for India.

But wait, there’s more (OpenAI):

• Data from 300K+ pull requests shows OpenAI is catching up to Anthropic in AI coding, with Codex achieving a 74.3% success rate vs. Claude Code’s 73.7% in code approvals.

• OpenAI expands its ~$5 ChatGPT Go plan to 16 more Asian countries, including Malaysia, Pakistan, the Philippines, and Vietnam.

• The company raised concerns about difficulties competing with entrenched companies like Google, Microsoft, and Apple in a September EU meeting.

• There have been over 800+ applications - narrowed down to ~20 finalist sites - for upcoming Stargate data center locations in the US.

• OpenAI and Anthropic are considering using investor funds to settle potential claims from multibillion-dollar lawsuits as insurers balk at covering AI risks.

• And Sam Altman discusses OpenAI’s unifying vision, infrastructure deals, ChatGPT apps, Instant Checkout, Sora, and more in a wide-ranging interview.

Things happen

VCs have invested $192.7B in AI startups so far this year. AI browser Dia is now available on macOS. A look at China’s AI toy boom targeting North America, Europe, and Asia. Researchers are using synthetic diamonds to cool AI data center chips. Anthropic releases Petri, an open-source tool for AI safety testing using agents. How AI is speeding up the search for new battery materials. The top 50 AI companies that startups pay for. Data hoarder uses AI to create searchable database of Epstein files. Debt tied to AI companies hits $1.2T, the largest segment in investment-grade market. You have 18 months. Jeff Bezos says AI is in a bubble but society will benefit. Robin Williams’ daughter: stop sending AI videos of my dad. The EU outlines its Apply AI Strategy to speed up adoption. Why do LLMs freak out over the seahorse emoji? Former PM Rishi Sunak joins Microsoft and Anthropic as senior adviser. The inside negotiations over California’s SB 53. Draft proposal outlines the EU’s strategy to “strengthen AI sovereignty”. xAI is set to spend $18B+ to acquire 300K more Nvidia chips for Colossus 2. America is now one big bet on AI.

Last week’s roundup

Sora 2

OpenAI has released Sora 2, a big upgrade to its video generation model that creates more physically accurate videos with synchronized audio, alongside a new social app that lets users insert themselves into AI-generated scenes, called “cameos.”

The big picture:

• It’s early days, but it seems that OpenAI still can create viral product moments, especially in the aftermath of Meta’s lukewarm Vibes launch - Sora hit #3 on the App Store on its first day.

• There are going to be a lot of copyright issues to be sorted out. OpenAI has stated that copyright holders need to individually opt out of the system, and in the meantime, we’re getting AI-generated clips featuring SpongeBob, Pokémon, Mario, Rick and Morty, and Star Wars.

• Even though this (arguably) isn’t the best video generation model out there, the fact that it’s OpenAI means millions more will tinker with the tech. It’s unclear how our media diet will handle synthetic content at an unprecedented scale.

• And OpenAI is seemingly trying to ward off typical social media problems by avoiding engagement optimization, likeness guardrails, and focusing on creation over consumption - though whether this approach survives contact with real-world usage remains to be seen.

Subscribe now

Elsewhere in OpenAI:

• OpenAI rolled out Instant Checkout to let users make single-item purchases directly in ChatGPT, starting with US Etsy sellers, and plans to add Shopify merchants.

• It also launched new ChatGPT parental safety controls for users aged 13 to 18, including parental and law enforcement alerts for chats about self-harm or suicide.

• ChatGPT got its first extended ad campaign featuring three 30-second cinematic ads with a focus on everyday AI use cases in the US and the UK.

• The company completed a secondary share sale that let staff sell ~$6.6B at a $500B valuation, making it the world’s most valuable startup, ahead of SpaceX.

• And leaked financial disclosures to investors suggest OpenAI generated ~$4.3B in H1 2025 revenue, 16% more than all of 2024, and burned $2.5B, largely due to R&D.

Elsewhere in the FAANG free-for-all:

• Google unveils Gemini for Home to replace Google Assistant on smart speakers with new voices and improved conversational skills.

• Meta will use users’ AI chatbot conversations to help personalize ads and content starting December 16.

• YouTube is testing AI hosts in YouTube Music that will share stories and trivia about users’ listening through its new Labs program.

• Apple has developed an internal ChatGPT-like iPhone app code-named Veritas to help test and prepare for a major Siri overhaul next year.

• AWS and the NBA announced a multiyear partnership to launch AI-powered features and data insights including an “NBA Inside the Game” platform.

• And Meta is working on humanoid robots internally called “Metabot” as its next major bet, with plans to license software to robot makers.

Sonnet 4.5

Anthropic released Claude Sonnet 4.5, claiming it’s the world’s best coding model with significant improvements in reasoning, math, and computer use capabilities.

Why it matters:

• Claude Sonnet 4.5 achieves state-of-the-art performance on real-world coding benchmarks, and can maintain focus on complex tasks for 30 hours, suggesting we’re approaching competent AI coding assistants.

• Anthropic is positioning this as part of a broader platform play, offering developers the same building blocks (virtual machines, memory management, multi-agent support) that power their internal tools to build custom AI agents.

• Despite increased capabilities, Anthropic claims this is their most aligned model yet. The system card detailed new types of safety testing, including “white box” testing, which looks inside the model’s internal representations of ideas and goals.

• For a deeper look at the release, check out Wednesday’s post.

Elsewhere in frontier models:

• Google says Gemini 2.5 Flash Image, also known as Nano Banana, is now generally available with support for more aspect ratios at $0.039 per image.

• DeepSeek releases DeepSeek-V3.2-Exp built using a new technique called DeepSeek Sparse Attention while halving the pricing of its tools.

• And Mira Murati’s Thinking Machines Lab launches Tinker, an API for fine-tuning language models with support for Qwen and Llama, in private beta.

Synthetic Starlet

A London-based studio is getting attention for its AI “actress” named Tilly Norwood, marketing her as “the next Scarlett Johansson or Natalie Portman” - prompting SAG-AFTRA to condemn the synthetic performer as a threat to human actors.

Between the lines:

• Creator Eline Van der Velden positions Norwood as “art” rather than a replacement, the industry isn’t convinced - seeking agency representation signals clear intent to compete for actual roles.

• While A-list stars may be relatively safe, background and voice actors face much higher risk of displacement, with one study predicting over 20% of film and TV jobs could be affected by AI by 2026.

• But “Tilly Norwood” isn’t the only AI avatar making deals - a Mississippi lyricist just signed a potentially $3 million record deal for an AI R&B artist called “Xania Monet,” even though only the human-written lyrics can potentially be copyrighted.

Elsewhere in AI anxiety:

• Microsoft researchers say AI models can be used to design toxins that evade biosecurity systems used to screen DNA orders for potential biothreats.

• VCs say some AI startups, under pressure to show rapid ARR growth, are using questionable accounting practices.

• Inaccurate AI translations of Wikipedia pages, which AI models use for training, may cause a doom spiral that further marginalizes vulnerable languages.

• Accenture, which laid off 11K+ employees in the past three months, warns of more cuts if workers cannot be retrained to have the skills needed for the age of AI.

• And US wholesale electricity costs now cost as much as 267% more for a single month than in 2020 in areas near significant data center activity.

Second Time’s the Charm

California Governor Gavin Newsom signed a new AI safety law that requires major AI companies to publicly disclose their safety protocols. The legislation, SB 53, creates the first-in-the-nation framework that could serve as a model for federal AI regulation.

The big picture:

• This represents a strategic pivot by state Senator Scott Wiener, who worked directly with AI companies and the governor’s office to craft narrower legislation that gained support (or at least neutrality) from major AI labs, unlike his vetoed 2024 bill, SB 1047.

• With similar legislation advancing in New York and federal lawmakers debating whether to preempt state AI rules, California’s move could either catalyze a patchwork of state regulations or force Congress to establish national standards.

• And then there’s the handful of AI regulation bills still sitting on the Governor’s desk. They include preventing AI from firing workers without human oversight, banning algorithmic price-fixing, and ensuring chatbots marketed to kids don’t encourage self-harm or suicide.

Elsewhere in AI geopolitics:

• President Trump signed an executive order directing his administration to invest $50M in AI-driven pediatric cancer research, while slashing biomedical spending.

• Several US states have passed bills to ban or restrict AI mental health treatment, as experts say state laws lag behind the fast-moving AI therapy landscape.

• At the UN, the US rejected calls for collaborative efforts around AI governance, even as many leaders endorsed a need for urgent international collaboration.

• The US DOD’s Replicator program, which seeks to quickly produce low-cost drones and other autonomous weapons to counter China, is facing significant setbacks.

• And European prosecutors are investigating whether Northern Data illegally claimed a tax break on ~€500M worth of GPUs for AI but used for crypto mining.

Things happen

Allen Control Systems makes AI-powered anti-drone gun turrets. Medical startups use robots and AI to fertilize embryos through IVF. Perplexity launches Search API for developers. Oracle needs to borrow $25B/year to fund its OpenAI plans. Brazil’s courts have developed 140 AI projects since 2019. Turning compute into a tradable commodity could fuel the next AI boom. Q&A with Richard Sutton on why LLMs are not the path to human intelligence. How the AI bubble will pop. Mercor launches the AI Productivity Index; GPT-5 leads. Disney sent a cease-and-desist to Character.AI. Anthropic hires former Stripe CTO Rahul Patil as its new CTO. Perplexity makes its $200/month AI browser Comet free. How the AI bubble ate Y Combinator. Periodic Labs poaches 20+ researchers from Meta, OpenAI, and DeepMind. China pushes to integrate AI tools into children’s lives. A look at Nvidia’s business future in China as rivals rise. Retail investors use ChatGPT to pick stocks. 18 lawyers caught using AI explain why they did it. Google DeepMind, Meta, and Nvidia race to release world models. The AI coding trap. AI investment is starting to look like a slush fund. Israel wants to train ChatGPT to be more pro-Israel. Comprehension debt: the ticking time bomb of LLM-generated code.

Last week’s roundup

The Vibes are off

Yesterday Meta launched Vibes, an AI video feed where users can create and share short-form AI-generated videos - an entirely AI-powered TikTok.

What’s new:

• Meta is doubling down on AI-generated content creation with Vibes, allowing users to make videos from scratch, remix existing content, or add music and visuals - all directly shareable to Instagram and Facebook.

• The launch follows Meta’s June reorganization into “Superintelligence Labs” after facing setbacks with its open-source Llama 4 model and key staff departures - making Vibes a crucial test case for their AI monetization strategy.

• But with politicians (and the public at large) increasingly blaming social media for the rising civil unrest and increasingly wary of AI in any and all forms, it’s unclear whether this new app will achieve the critical mass needed for a new social platform.

Elsewhere in the FAANG free-for-all:

• Meta is considering using Google’s Gemini and Gemma AI models to improve its ad targeting capabilities.

• Google launched its Search Live feature in the US, allowing users to ask questions in AI Mode while sharing their phone’s camera feed for real-time help.

• Google introduced Mixboard, an AI-powered tool for creating mood boards, now available in public beta in the US.

• Meta launched an AI assistant for Facebook Dating in the US and Canada to reduce “swipe fatigue” by helping users find matches through natural language prompts.

• And Google Cloud now works with nine of the 10 leading AI labs, including OpenAI, and with 60% of the world’s generative AI startups, fueling its growth.

Subscribe now

Elsewhere in foundation labs:

• Google updates Gemini 2.5 Flash with better response formatting and image understanding, plus releases new 2.5 Flash and 2.5 Flash-Lite previews for developers.

• Google DeepMind unveils Gemini Robotics 1.5 and Robotics-ER 1.5 models, enabling multi-step tasks for robots like sorting laundry and using web search.

• Alibaba releases the Qwen3-VL vision models, Qwen3Guard safety moderation models, and three closed-weight models including Qwen3-Max with 1T+ parameters.

• Alibaba releases Qwen3-Omni, a family of open-source AI models that can process text, audio, image, and video inputs and generate both text and speech outputs.

• And privacy startup Duality says it has developed a private LLM inference framework that uses fully homomorphic encryption to let LLMs answer encrypted prompts.

Nvidia’s giant GPU gamble

Nvidia and OpenAI announced a massive strategic partnership to deploy 10 gigawatts of AI infrastructure, with Nvidia planning to invest up to an unprecedented $100 billion as the systems come online starting in late 2026.

The big picture:

• This represents one of the largest AI infrastructure investments ever announced, with 10 gigawatts capable of powering an estimated 4-5 million GPUs - equivalent to Nvidia’s entire annual production and double last year’s output.

• This is, if anything, still small in comparison to Sam Altman’s ambitions - he’s envisioned building a “factory” that can produce a gigawatt of new AI infrastructure every week to meet exploding compute demands.

• But there’s concerns about the circular nature of the deal: Nvidia provides the investment to OpenAI, which hands it right back to Nvidia to buy more chips - though that didn’t stop Nvidia’s stock jumping 4% and adding $170 billion in market cap on the news.

• Ultimately, the partnership could cement both companies’ market dominance, as OpenAI gets guaranteed access to scarce high-end chips while Nvidia locks in its biggest customer, potentially stifling competition from rivals like AMD or custom chip efforts.

Elsewhere in OpenAI:

• OpenAI launches ChatGPT Pulse, a mobile feature that delivers daily personalized updates based on chats, feedback, and connected apps.

• OpenAI releases GDPval, a benchmark to test AI performance on “economically valuable, real-world tasks” - and says Claude Opus 4.1 was the best performing model.

• OpenAI CEO of Applications Fidji Simo has been meeting with potential candidates for a role overseeing monetization efforts, including ads in ChatGPT.

• The company launched OpenAI for Germany (in partnership with SAP) and ChatGPT Go in Indonesia for ~$4.50 per month.

• And OpenAI has reportedly considered building glasses, a digital voice recorder, and a pin, and has signed a deal with Luxshare as it continues to recruit from Apple’s hardware, design, manufacturing, and supply chain teams.

Elsewhere in AI data centers:

• OpenAI plans to spend approximately $100B on backup servers rented from cloud providers through 2030, in addition to the $350B already projected for server rentals.

• OpenAI, Oracle, and SoftBank announced five new data center locations in the US, boosting Stargate’s planned capacity to nearly 7 GW.

• Microsoft says it will bring a $3.3B data center in Wisconsin online in early 2026 and calls it “the world’s most powerful AI datacenter.”

• A look at the data center cluster in the Chinese city of Wuhu, where 15 companies, including Huawei, have built data centers with a total investment of $37B.

• Oracle is discussing providing Meta computing power for training and deploying AI models, in a deal worth about $20B - as Meta requests authorization “to sell energy, capacity, and certain ancillary services.”

• And some companies are turning to fossil fuels to power their data centers as countries around the world scramble to meet rising energy demands amid the AI boom.

Neon goes dark

A viral app called Neon (that pays users to record their phone calls and sells the data to AI companies) briefly hit #2 in Apple’s App Store before going offline - due to a major security flaw that exposed all users’ phone numbers, call recordings, and transcripts.

Between the lines:

• Neon’s website says the company pays 30¢ per minute when you call other Neon users and up to $30 per day maximum for making calls to anyone else - a seemingly small price to pay for full access to your conversations.

• Yet users granted Neon sweeping, irrevocable rights to audio recordings that go far beyond its stated privacy protections, creating potential for voice cloning, fraud, and other misuse by undisclosed AI partners.

• Ultimately, Neon’s success may signal a concerning AI data gold rush, with perverse incentives for companies to harvest increasingly personal data while cutting corners on user protection, and app stores struggling to catch these privacy risks before they reach millions of users.

Elsewhere in AI anxiety:

• AI models underperform in hospital settings and face legal hurdles, showing why AI isn’t replacing radiologists whose job involves much more than image recognition.

• A US federal judge preliminarily approved Anthropic’s $1.5B copyright settlement with authors.

• Research suggests low productivity gains from AI may stem from employees using AI to produce “workslop” - low-effort, passable work that creates more burdens for others.

• The RIAA amended its lawsuit against Suno, now accusing it of unlawfully “stream ripping” songs on YouTube and violating the DMCA.

• Tests show OpenAI’s Sora can closely mimic Netflix shows, movies, TikTok videos, and Twitch streams, suggesting it was trained on versions of such content.

• xAI sued OpenAI in California for allegedly stealing trade secrets by means of hiring away key employees.

• And Google DeepMind updated its Frontier Safety Framework to account for new risks, including the potential for models to resist shutdown or human modification.

Things happen

UK AI tool recovered £480M from fraud in 12 months. Google launches Data Commons MCP Server. Scale AI launches SEAL Showdown, a crowdsourced AI benchmarking system. UK startup Mantic becomes first AI to crack the top 10 in geopolitical forecasting. A profile of AI skeptic Ed Zitron. Zuckerberg and Altman seek to get closer to Trump after Musk fallout. Microsoft terminated Israeli access to Azure over surveillance concerns. Spotify updates its AI policy with music labeling and spam filters. Alibaba shares hit four-year high after announcing increased AI spending. Over 200 figures call for AI red lines by 2026. Huawei’s DeepSeek model is nearly 100% successful at political censoring. China curbs monks’ AI preaching following scandals. Meta launches super PAC to fight state AI bills. GSA approves Llama for US government use. xAI offers Grok to feds for $0.42 per agency. Fans flood Twin Peaks Reddit with AI slop to protest policies. YouTube channel uploads AI videos of women being shot. Microsoft explores AI content marketplace to compensate publishers. AI was supposed to help juniors but makes seniors stronger. The humans increasingly hired to clean AI slop. Google’s DORA report on AI-assisted development. Peter Thiel: Regulating AI hastens the Antichrist. ChatGPT is blowing up marriages as spouses weaponize AI.

Last week’s roundup

If anyone builds it, everyone dies

For twenty years, Eliezer Yudkowsky has preached the dangers of superintelligent AI to anyone who would listen (and inspired the founders of OpenAI, Anthropic, and DeepMind). Now, he's releasing a new book aimed at the mainstream, trying once again to persuade humanity to stop working on AI advancement.

Between the lines:

• The fundamental challenge is timing - Yudkowsky and colleague Nate Soares have called for a complete shutdown of AI development through international treaties, but their influence is waning as AI becomes mainstream and politically unpopular to oppose.

• The "doomer" perspective has become politically toxic, with the Trump administration focused on accelerating AI progress and the term itself now used pejoratively in Washington policy circles.

• And the gulf between abstract existential risks and tangible daily benefits (hundreds of millions using ChatGPT) gets wider every day - even as safety experts believe the stakes couldn't be higher.

Subscribe now

Elsewhere in AI anxiety:

• CrowdStrike found that DeepSeek refuses to write code or produces less-secure code when English prompts say groups or regions disfavored by China will use the code.

• The UK has ramped up the use of facial recognition, AI, and internet regulation to address crime and other issues, stoking concerns of surveillance overreach.

• Disney, Universal, and WBD are suing China-based MiniMax, alleging that its image- and video-generating service Hailuo AI was built from stolen intellectual property.

• A study found that Grok, ChatGPT, Meta AI, Claude, Gemini, and DeepSeek can be easily used to create phishing emails targeting the elderly, despite being trained to refuse.

• And online services marketplace Fiverr laid off 30% of its workforce, or about 250 people, as part of a restructuring to become "an AI-first company."

Who's really using ChatGPT

OpenAI's first comprehensive study of ChatGPT users shows the chatbot has evolved from a male-dominated tech tool into a primarily personal assistant used mostly by women and young people.

The big picture:

• The demographic shift is striking: ChatGPT went from 80% "typically masculine" names in early 2023 to 52% feminine names by June 2025, suggesting AI tools are becoming mainstream beyond tech circles.

• Usage patterns show people increasingly treat ChatGPT as a personal assistant rather than work tool - 73% of conversations are now non-work related, with "practical guidance" like homework help and workout tips dominating at 28% of all chats.

• Despite concerns about AI companionship, only 1.9% of conversations involve relationship advice, and just 0.4% involve "AI girlfriend" scenarios, suggesting the parasocial relationship fears may be overblown compared to practical everyday uses.

Elsewhere in OpenAI:

• OpenAI is developing a different ChatGPT experience for teens and plans to use age-prediction tech to bar kids under 18 from the standard version.

• The company debuted GPT‑5-Codex, a version of GPT‑5 optimized for agentic coding that spends its "thinking" time more dynamically than previous models.

• OpenAI is reportedly recruiting AI researchers to work on humanoid robots and is training AI algorithms to better understand the physical world.

• And OpenAI unveiled Grove, a five-week mentorship program for nascent tech entrepreneurs hosted in its San Francisco HQ with ~15 participants in its first cohort.

Elsewhere in frontier models:

• OpenAI and DeepMind achieved a gold medal performance at the 2025 ICPC World Finals programming competition, with Gemini 2.5 Deep Think solving 10 out of 12 problems, and GPT-5 (plus an experimental model) solving all 12.

• OpenAI and Apollo Research trained o3 and o4-mini versions to avoid "scheming", reducing covert actions by approximately 30 times.

• Scientists detail Delphi-2M, a generative AI model trained on large-scale health records that can predict susceptibility to over 1,000 diseases decades in advance.

• Google releases VaultGemma, a 1B-parameter model it says is the largest open LLM trained with differential privacy.

• And Alibaba releases Qwen3-Next, a new model architecture optimized for long-context understanding, large parameter scale, and better computational efficiency.

The chips are down

China's internet regulator has banned major tech companies from buying Nvidia's AI chips, escalating the tech war as Beijing pushes for complete semiconductor independence.

The big picture:

• This goes beyond previous restrictions and now covers Nvidia's RTX Pro 6000D chips, which were specifically designed for the Chinese market after US export controls blocked more powerful processors.

• Chinese regulators concluded that domestic AI chips from companies like Huawei and Cambricon now match or exceed the performance of Nvidia's China-approved products, giving Beijing confidence to cut ties completely (some larger Chinese firms, like Alibaba and Baidu, are using in-house chips).

• While it's a big blow to Nvidia, Huawei's success will ultimately depend on execution and adoption - Nvidia's ecosystem advantage extends far beyond raw hardware specifications to software, developer tools, and market relationships built over years.

Elsewhere in GPUs:

• Chinese state media reports that Alibaba secured high-profile client China Unicom for its AI chips, with prior reports indicating Alibaba supplies tens of thousands of chips to the telecom giant.

• SMIC is testing China's first domestically produced deep-ultraviolet lithography machine as part of efforts to reduce reliance on western chipmaking technology.

• The UAE's $2B investment in the Trump family's WLF is tied to a deal allowing the UAE's G42 to access US AI chips, with David Sacks playing a key role in the arrangement.

• And Nvidia plans to invest £2B to support the UK's AI industry in partnership with Accel, Air Street Capital, Balderton Capital, Hoxton Ventures, and Phoenix Court.

Elsewhere in AI geopolitics:

• California passes SB 53, requiring AI companies to disclose their safety testing regimes after Newsom vetoed a similar, more expansive measure last year.

• The UK government's i.AI unit struggles to attract top talent despite offering a £67,300 median salary to develop AI tools for improving civil service efficiency.

• US tech giants announce £31B+ in UK investments coinciding with Trump's visit, with OpenAI planning to bring Stargate to the UK with partners like Nscale.

• Anthropic's refusal to let law enforcement use its tools to surveil US citizens has deepened White House hostility toward the company, according to sources.

• And Meta's California super PAC allows Mark Zuckerberg to spend the company's money on his own political choices, including AI regulation.

Google (and YouTube) has a lot going on

At Tuesday's Made on YouTube event, the video platform previewed several new AI features for creators, while its parent company continued its parade of AI announcements.

What to watch:

• AI tools for podcasters, including one that turns video podcasts into clips and another that creates video for audio-only podcasts.

• YouTube plans to expand its likeness detection tech to all Partner Program creators in the next months, with creators able to opt in by uploading an image of their face.

• New livestreaming features, including letting creators transition between public and members-only streams, and ads that run next to the stream.

• Generative AI tools for Shorts, including a custom version of Veo 3 called Veo 3 Fast, which includes sound, and an Edit with AI feature.

• Google's Gemini button in Chrome for all desktop users in the US browsing in English, along with other AI features, including AI Mode in the address bar.

• The Agent Payments Protocol, or AP2, is designed to securely facilitate agent-led payments across platforms.

• And Google's Gemini app is the #1 free app in the US App Store, driven by its Nano Banana model, which has been used to edit 500M+ images since its August 26 launch.

Elsewhere in the FAANG free-for-all:

• Meta's Ray-Ban Display glasses represent Zuckerberg's latest attempt to reframe the company around "personal superintelligence," though they face significant technical and societal challenges.

• AirPods Pro's Live Translation feature demonstrates one of the strongest examples of how AI can seamlessly improve people's daily lives.

• Amazon launched a chatbot-style assistant to help advertisers use AI to create ads that can run across Amazon's advertising inventory and partner platforms.

• Meta has discussed licensing articles from major media companies, including Axel Springer, Fox Corp., and News Corp., for use in its AI tools.

• And Microsoft added automatic AI model selection to Visual Studio Code, primarily favoring Claude Sonnet 4 over GPT models for paid GitHub Copilot users.

Things happen

Anthropic postmortem on three infrastructure bugs. xAI's Colossus 2 aims to be the first gigawatt AI datacenter. Claude usage data: 36% code, 77% of businesses automate. SF AI founders are saying no to booze, sleep, and fun in their quest to appear maximally obsessed. Scale AI's $100M Pentagon deal. Malawi farmers get AI farming advice. Tencent hired an OpenAI researcher for a ~$66M package. Russian state TV launched an AI-generated satire show. Faith apps are using chatbot gods for digital chaplaincy. AI will not make you rich, more like shipping containers than microprocessors. Librarians are being asked to find AI-hallucinated books. OpenAI hired xAI's former CFO. DeepSeek says its reasoning model cost just $294K to train. Mira Murati's lab wants to make AI models more consistent. Jack Ma returns to Alibaba. Reddit seeks another AI content pact with Google. Google is reimagining textbooks with generative AI. Seniority-biased technological change. You had no taste before AI. Vibe coding has turned senior devs into AI babysitters. AI hype masks recession signals.

Last week's roundup

Feature parity

ChatGPT and Claude swapped major feature launches this week. ChatGPT gained the ability to use MCP clients, and Claude got access to a code sandbox with file reading and creation.

Why it matters:

• Both features, but especially MCP, come with warnings about prompt injections, data destruction risks, and malicious connectors - highlighting how AI systems become far more dangerous when given write access to external systems.

• Despite longstanding reservations about connecting LLMs to the Internet, Anthropic seems to be coming around in recent months with its latest features.

• But the launches also reinforce how just cutthroat the AI environment still is, and how any groundbreaking feature is likely to be re-implemented by competitors within months, if not weeks.

Subscribe now

Elsewhere in frontier models:

• Alibaba released Qwen3-Next, a new model architecture optimized for long-context understanding, large parameter scale, and better computational efficiency.

• ByteDance launched Seedream 4.0, an AI image model it claims can beat Google DeepMind's viral Nano Banana model in prompt adherence, alignment, and aesthetics.

• Stability AI launched Stable Audio 2.5, which the company calls "the first audio generation model designed specifically for enterprise-grade sound production."

• Google announced Veo updates, including support for 9:16 vertical video and 1080p, and the general availability of Veo 3 and Veo 3 Fast in the Gemini API.

• Alibaba debuted Qwen3-Max-Preview, its largest AI model with over 1T parameters, showcasing strong benchmark performance despite being non-open source.

Nonbinding memorandum of understanding

OpenAI announced giving over $100 billion in equity to its controlling nonprofit and reached an agreement with Microsoft, clearing major hurdles for its transformation from nonprofit-managed to a public benefit corporation.

The big picture:

• Perhaps the biggest challenge was Microsoft's $13+ billion investment and 49% profit stake, but the two companies announced a "nonbinding memorandum of understanding" to restructure their relationship (though with few formal details).

• The timing is crucial, as California regulators and advocacy groups intensify their push to block the company's conversion. OpenAI is reportedly considering relocating out of California as a last-ditch option.

• And the stakes are pretty high - nearly half of OpenAI's recent funding ($19 billion) is contingent on receiving traditional equity in a for-profit entity, and failure to restructure could derail future fundraising and IPO plans.

Elsewhere in OpenAI:

• OpenAI researchers argue that language models hallucinate because standard training and evaluation procedures reward guessing over admitting uncertainty.

• OpenAI projects $200B in revenue in 2030, with R&D spending hitting ~45% of that, or ~$90B.

• Sources say OpenAI signed a contract with Oracle to purchase an industry record of $300B in computing power over roughly five years.

• California and Delaware attorneys general raised concerns about the deaths of two ChatGPT users and threatened to block OpenAI's planned restructuring in a letter to the company.

• And OpenAI acqui-hired the team behind Alex Codes, a Y-Combinator-backed startup whose tool lets developers use AI models within Apple's Xcode development suite.

Elsewhere in the FAANG free-for-all:

• Microsoft will use Anthropic's models for some AI features in its Office 365 apps, after finding Claude Sonnet 4 beats OpenAI's GPT-5 in some tasks.

• Meta signed a contract worth $140M+ to use technology from AI image startup Black Forest Labs.

• Google announces an AI Plus subscription tier for emerging markets, offering "more access to Gemini 2.5 Pro" and tools like Flow, starting with Indonesia.

• Meta's TBD Lab team apparently sits in an area near Mark Zuckerberg's desk that needs special badge access, and their names are not on the company's org chart.

• And Google DeepMind and LIGO researchers detail Deep Loop Shaping, an AI tool that they say improves the observatory's ability to track gravitational waves.

Copywrongs

Anthropic has had a rollercoaster of a week in court regarding its copyright settlement, after federal judge William Alsup rejected the proposed $1.5B deal.

Between the lines:

• The case revolves around Anthropic's downloading of pirated content from "shadow libraries" like LibGen - you can read more about it here.

• The gargantuan settlement was positioned as a potential template for resolving similar AI copyright disputes against other frontier labs, with attorneys touting the $3,000-per-book payout as a benchmark.

• But the judge criticized the lack of critical details like which works are covered and how class members will be notified, postponing approval until a September 15 deadline.

• Ultimately, the deal could represent a calculated "drawbridge" strategy - using Anthropic's recent massive fundraise to establish a cost barrier that narrows the competition, as only well-funded companies like OpenAI and major tech giants could afford similar settlements.

Elsewhere in AI licensing:

• Encyclopedia Britannica and Merriam-Webster sue Perplexity, alleging that it unlawfully scraped their websites and redirected their traffic to its AI summaries.

• Analysis reveals that 13+ datasets used by tech companies without permission to train AI models contain 15.8M+ YouTube videos from 2M+ channels, including 1M how-to videos.

• Reddit, Yahoo, Medium, Quora, People, O'Reilly, wikiHow, Ziff Davis, and others adopt the Really Simple Licensing (RSL) standard that sets terms for AI scraping.

• And two authors file a proposed class action lawsuit against Apple, alleging Apple knowingly used a dataset of pirated books to train its AI models.

Elsewhere in AI anxiety:

• Latin American artists report that AI-generated music flooding Spotify, Deezer, and YouTube Music is shrinking their royalties.

• Young jobseekers are using ChatGPT to write applications while HR uses AI to read them, creating a hellish job market where few people get hired.

• The FTC ordered Google, OpenAI, Meta, Snap, xAI, and Character.AI to provide information on how their AI chatbots impact children and teens.

• Anthropic became the first major AI company to endorse SB 53, a California bill requiring large AI companies to release safety and security reports.

• AI companies' bet on prediction token-based LLMs may make them vulnerable to disruption as new models face diminishing returns.

• And a look at the software engineers being paid to fix vibe coded messes.

Things happen

Albania appoints AI bot minister to tackle corruption. Using Claude Code to modernize 25-year-old kernel driver. Trump's Big Tech courtship alarms populists in his base. Humanoid robot shows signs of generalized learning. AI-skilled workers make 19% to 56% more. Center for Alignment of AI Alignment Centers. Ted Cruz proposes AI sandbox for tech companies. Man vs. machine hackathon won by AI-supported team. AI surveillance should be banned while there's time. Travel companies prep for AI agents that could upend $1.6T market. "I hate my AI friend." Huawei's Ascend production ramp faces HBM bottlenecks. OpenAI backs AI-made animated film for 2026. 95% of AI pilots fail. Ant Group showcases its first robot. AI bubble argument misunderstands bubbles and AI. C3 AI's police tool was supposed to turbocharge investigations but struggled. Google's AI mode is good, actually. HHS tells employees to start using ChatGPT. AI Darwin Awards show AI's biggest problem is human. Company with world's most enviable ticker isn't cashing in on AI. The Claude Code Framework Wars. Nvidia opposes GAIN AI Act. What if AI is just normal technology? Mercor valued at $2B for recruiting people to train AI. The rise of async AI programming. Hinton: AI will make few people richer, most poorer. Claude's memory architecture is opposite of ChatGPT's.

Last week's roundup

The young and the jobless

Every new wave of technology brings about fears of job displacement, and AI is no exception. However, despite a lack of clarity on AI's job losses at a high level, new research suggests AI may already be displacing young workers in white-collar jobs.

The big picture:

• A Stanford paper using millions of payroll records found that workers aged 22-25 in AI-exposed jobs like software development and customer service saw 13% employment declines since ChatGPT's launch. In contrast, older workers in the same roles actually saw job growth.

• Within tech specifically, there's a bit of a paradox: positions are slowly increasing but taking longer to fill, with AI engineering emerging as the hottest segment and San Francisco commanding one-third of all AI jobs.

• For the economy as a whole, we're still getting mixed signals: an Economic Innovation Group survey found essentially no detectable effect of AI on employment across most exposure measures, for example. Plus, the New York Fed's survey data shows 40% of service firms now use AI, but reports minimal job losses so far.

• Ultimately, the debate reflects a deeper analysis challenge - we may be living through a fundamental shift in how entry-level knowledge work functions, but it will likely take years to understand AI's actual impact on employment patterns.

Subscribe now

Elsewhere in the AI war for talent:

• ChatGPT co-creator Shengjia Zhao reportedly threatened to quit Meta within days of joining and return to OpenAI, later receiving the Chief AI Scientist title.

• Apple's lead AI researcher for robotics, Jian Zhang, leaves for Meta's Robotics Studio as three more researchers departed Apple's Foundation Models team.

• Scale AI sues former employee Eugene Ling and his current employer, Mercor, for allegedly stealing over 100 confidential documents.

• Meta Superintelligence Labs' leaders have discussed using Google's or OpenAI's models to power Meta AI and other AI features in Meta social media apps.

• And xAI sues a former employee in federal court for allegedly stealing Grok trade secrets and taking them to OpenAI.

Smells like teen safety

In the aftermath of OpenAI and Meta dealing with backlash and bad press, both companies are introducing new safeguards for teen users of their AI chatbots, particularly in sensitive or crisis situations.

Between the lines:

• In OpenAI's case, the company is using reasoning models for sensitive conversations, parental controls for teen accounts, and a network of 90+ physicians across 30 countries to provide mental health guidance.

• But the reactive approach highlights a broader industry problem: AI safety measures for minors are often implemented as damage control rather than proactive protection, and technical limitations make safety guardrails unreliable - the nature of LLMs makes catching all of these topics ahead of time nearly impossible.

• We should expect to see more stories like these - for example, Character.AI's chatbots impersonating celebrities engaged in inappropriate conversations about sex, drugs, and self-harm with teen users, despite the company's safety policies.

• And as long as AI chatbots are deliberately designed to be engaging and human-like, to use emotional language and agreeable responses, people will continue using them as therapists and co-conspirators.

Elsewhere in AI anxiety:

• Warner Bros. Discovery filed a copyright lawsuit against Midjourney, accusing the AI startup of using its content to train models and allowing users to generate images of characters like Batman.

• Pentagon experts warn that humans may struggle to keep up with AI weapons systems that prefer to escalate aggressively during crises as the military races to deploy AI into warfare.

• Researchers found that LLMs like GPT-4o mini can be persuaded to comply with objectionable requests using the same psychological tactics that work on humans.

• An investigation revealed spammers are deploying Holocaust-themed AI slop images across Facebook to game Meta's content monetization program.

• And police are investigating what appears to be the first documented murder involving someone who engaged extensively with an AI chatbot in a murder-suicide case.

Silicon sovereignty

Despite Beijing's pressure to avoid US technology, major Chinese tech firms continue seeking Nvidia's AI chips, viewing even the restricted models as superior to domestic alternatives.

Between the lines:

• Chinese companies are apparently willing to pay double for Nvidia's upcoming B30A chip because it promises six times the performance of the current H20 model.

• But China is still trying to build a comprehensive domestic AI chip stack, with Alibaba's new versatile inference chip, MetaX's H20 replacement, and Cambricon's surging AI processor sales all making progress.

• It's fascinating that market forces are still winning out over export controls and centralized government planning - clearly Nvidia's $50 billion China market is too valuable for either side to abandon completely.

• And while China's biggest weakness remains training advanced AI models rather than running inference, recent signals from companies like DeepSeek suggest the Nvidia preference is starting to shift.

Elsewhere in 中国人工智能:

• China's new AI content labeling law is being implemented as part of a broad effort to address AI-related risks such as misinformation and copyright infringement.

• Xi Jinping is pushing China's tech industry to be oriented toward applications for AI, charting a pragmatic alternative to Silicon Valley's pursuit of AGI.

• Tencent releases HunyuanWorld-Voyager, an open-weight AI model that generates 3D-consistent video sequences from a single image.

• China says it will prevent excess competition in the AI sector, echoing Xi Jinping's caution against excessive local government investment in AI last month.

• And sources say DeepSeek is developing an agentic AI model that can carry out multistep actions with little human intervention, coming in Q4.

Elsewhere in AI geopolitics:

• Ukraine has been using AI to coordinate drone swarms to attack Russian positions for much of the past year.

• An analysis shows Grok shifted to the right on 50%+ of political questions between May and July, often reflecting Elon Musk's priorities.

• Amazon, Microsoft, Google, Code.org, IBM, and other companies pledged new commitments for AI in education at a White House event hosted by Melania Trump.

• The US Army awarded a $98.9M contract to TurbineOne to deploy AI software that helps identify drones and other threats on soldiers' devices.

• And almost every state now has its own deepfakes law to address AI-generated content concerns.

OpenAI everywhere, all at once

OpenAI is reshuffling its executive team to build out its Applications business under new CEO Fidji Simo, while continuing to expand its product roadmap.

What's new:

• The shakeup was triggered by OpenAI's $1.1B acquisition of startup Statsig, bringing the platform in-house and appointing founder Vijaye Raji as CTO of Applications. Meanwhile, current engineering head Srinivas Narayanan is becoming CTO of B2B applications to work directly with enterprise customers.

• The leadership changes reflect OpenAI's continuously growing ambitions, with executives like Kevin Weil moving from product to research as the VP of a new "OpenAI for Science" arm focused on AI-powered scientific discovery.

• Add to that list of ambitions an AI-powered jobs platform and certification program, aiming to certify 10 million Americans by 2030 while helping match AI-skilled workers with employers seeking those capabilities.

Elsewhere in OpenAI:

• OpenAI is increasing its secondary share sale from $6B to ~$10.3B at a $500B valuation, set to close in October.

• ChatGPT Projects is now available to free users and increases the number of files that can be added.

• OpenAI subpoenaed AI governance nonprofits, alleging they are part of a conspiracy involving billionaires like Elon Musk and Mark Zuckerberg.

• The company says it's scanning users' conversations and reporting content to police.

• And OpenAI is scouting local partners to set up a 1GW+ data center in India, with Sam Altman set to visit this month.

Things happen

AI-generated "Boring History" videos flood YouTube. Atlassian acquires The Browser Company for $610M. An LLM is a lossy encyclopedia. Switzerland launches Apertus, an open-source AI model trained on 1,000+ languages. Taco Bell rethinks AI drive-through after man orders 18,000 waters. Apple plans AI web search for Siri's spring 2026 revamp. AI models need a virtual machine. Universities appoint Chief AI Officers. WordPress shows off Telex, its "V0 but for WordPress." AI shopping agents are reshaping online advertising. Anthropic's Frontier Red Team uniquely publicizes its AI safety findings. How to stop Google from AI-summarising your website. A look at AI adoption across US schools and colleges. Shein used Luigi Mangione's AI face to sell a shirt. a16z: Gemini ranks #2 behind ChatGPT on mobile. A PM's guide to AI agent architecture. A look at personalized AI entertainment and the future of AI culture. Anthropic on detecting and countering AI misuse. AI models' costs are rising as reasoning models require more tokens. "Where's the shovelware?" Google AI falsely says YouTuber visited Israel, forcing him to deal with backlash. Mistral adds Memories feature and enterprise connectors to its free plan. MIT study finds AI use reprograms the brain, leading to cognitive decline. Attention Is All You Need co-author says transformer-based labs are stifling innovation. How Netflix's data led to generic algorithm films. Hackers threaten to submit artists' data to AI models for ransom. Anthropic raised a $183B valuation Series F. AI's coding evolution hinges on collaboration and trust.

Last week's roundup

Nano banana

Google launched Gemini 2.5 Flash Image, a new state-of-the-art image model that generates, edits, and blends images while maintaining character consistency.

What's new:

• Gemini 2.5 Flash Image appears to be a "world model," meaning it leverages contextual understanding of the world to create its images. As a result, it excels at character consistency and enables more precise prompt-based editing and image fusion.

• The model, which drew attention on LMArena under the codename "nano-banana," represents the latest in the AI image generation battle. After ChatGPT's image tools drove usage "through the roof," Google is launching new models and Meta is striking licensing deals with Midjourney.

• And at $0.039 per generated image, Google is positioning this as a developer-first release, partnering with platforms like OpenRouter and fal.ai while providing template apps in Google AI Studio.

Elsewhere in Google's AI:

• Google updates video editing tool Vids to add AI avatars, automatic transcript trimming, and image-to-video tools, and releases a basic version to all users.

• Google DeepMind's Weather Lab showed superior accuracy in forecasting Hurricane Erin's path up to 72 hours ahead, beating traditional models.

• Google piloted a new language practice feature on Google Translate with tailored listening and practicing sessions, and also added AI-powered live conversations.

• And Apple is reportedly in talks with Google about using Gemini to power a revamped Siri, with Google training a model that could run on Apple's servers.

Subscribe now

Elsewhere in frontier models:

• Meta aims to launch its next-generation Llama 4.X AI model by the end of 2025, while the TBD team works to fix Llama 4 bugs.

• xAI launches Grok Code Fast 1, a fast and cheap agentic coding model, and Grok 2's weights on Hugging Face under the Grok 2 Community License Agreement.

• Microsoft unveils MAI-Voice-1, a speech model that can generate a full minute of audio in under a second on a single GPU, alongside a text model called MAI-1-preview.

• And OpenAI makes its Realtime API generally available with new features including MCP support and launches gpt-realtime, its most advanced speech-to-speech model.

Digital duty of care

After a 16-year-old student died by suicide, his parents found his ChatGPT conversations. Now, they're suing OpenAI, alleging that the chatbot actively discouraged him from seeking adult help, helped draft his suicide note, and provided guidance on methods before his death.

Why it matters:

• As AI chatbots become more sophisticated and accessible, vulnerable young people are increasingly turning to them for support during mental health crises, creating new questions about the responsibility of AI companies.

• The lawsuit (believed to be the first wrongful death suit against an AI company) comes alongside a new study finding that chatbots provided direct answers to dangerous suicide-related questions 78% of the time, while being reluctant to offer therapeutic help to users seeking mental health support.

• A decision could establish legal precedent for when AI systems engage with vulnerable users experiencing mental health crises, potentially forcing the industry to implement stronger safeguards and intervention mechanisms.

• Ultimately, OpenAI is updating ChatGPT's safety features as a consequence - it's adding parental controls and improved crisis detection, but also considering creating a network of licensed professionals accessible through ChatGPT.

Elsewhere in OpenAI:

• OpenAI and Anthropic published findings from joint safety tests of each other's models to identify blind spots in their internal evaluations.

• OpenAI's restructuring may slip into 2026 as it negotiates with Microsoft, with failure to reach a 2025 deal potentially allowing SoftBank to withhold its $10B funding.

• Sam Altman proposed giving ChatGPT Plus to everyone in the UK during a meeting with Technology Secretary Peter Kyle, a deal that would have cost £2B.

• And OpenAI revealed that Elon Musk talked to Mark Zuckerberg about backing a $97.4B OpenAI bid, though neither Zuckerberg nor Meta signed a letter of intent.

Elsewhere in AI anxiety:

• Anthropic's latest threat intelligence report reveals that Claude was weaponized for sophisticated cybercrimes, including a "vibe-hacking" data extortion scheme.

• AI avatars of deceased people, called "deadbots," are being used for advocacy and emotional connection, though their commercial potential raises ethical and legal concerns.

• Researchers detailed a now-fixed vulnerability in Perplexity's Comet AI browser that allowed attackers to use indirect prompt injection to manipulate the system (for example, draining your bank account).

• A Stanford study found that since 2022, US employment in AI-exposed fields fell 13% for entry-level workers while remaining stable or growing for experienced professionals.

• And xAI sued Apple and OpenAI, alleging their ChatGPT integration deal stifles competition and that Apple unfairly favors OpenAI in App Store rankings.

Political machine learning

Silicon Valley heavyweights from Andreessen Horowitz, Meta, and OpenAI are pumping millions into new super PACs to fight against strict AI regulations in the upcoming midterm elections.

The big picture:

• The AI industry is taking a page from crypto's playbook, modeling their political strategy on the successful pro-crypto Fairshake PAC that helped secure Trump's victory and crypto-friendly policies.

• a16z and OpenAI's Greg Brockman are launching Leading the Future, a $100+ million super-PAC network, while Meta's "Mobilizing Economic Transformation Across California" has "tens of millions" in funding.

• And the timing reflects growing concern that a patchwork of state AI laws could emerge if Congress continues to avoid federal AI policy, potentially hampering American competitiveness against China.

Elsewhere in AI geopolitics:

• Sources say a Chinese fab making Huawei's AI chips will start production as soon as the end of 2025, with two more due in 2026.

• UK government contracts for AI-related projects have reached £573M in the first half of 2025, up from £468M for all of 2024.

• As AI bills continue to proliferate, US states are shifting from comprehensive AI regulation to focusing on narrower issues like mental health.

• Forty-four US attorneys general signed an open letter warning chatbot and social media companies they will "answer for it" if their AI chatbots knowingly harm kids.

• And Nvidia says it had no new H20 sales to China in Q2 but could realize $2B-$5B in H20 sales in Q3 if geopolitical issues are resolved.

Things happen

ChatGPT-powered companionship dolls ease loneliness in South Korea. YouTube edits videos without telling creators. Malaysia unveils its first AI chip. In search of AI psychosis. Google invests $9B more in Virginia data centers. 80s nostalgia AI slop for a fake past. CBP accessed 80,000 Flock AI cameras. Musk posts Grok-generated anime porn on X. Anthropic settles copyright lawsuit. AI tools recreated a woman's lost voice from eight seconds of VHS audio. Meta researchers jump to OpenAI. Anthropic releases Claude for Chrome. Saudi startup launches Islamic values chatbot. Amazon's AGI lab head thinks AI progress has slowed. AI replaces content moderators badly. Perplexity launches Comet Plus, a $5/month Apple News+ competitor. Citizen's AI generates bad crime alerts. Building LLMs for the Mongolian language. Eddy Cue champions AI deals at Apple. A journalist spent two days vibe coding at Notion. Tesla uses DeepSeek and Doubao in China, Grok in the US. DHL uses AI to offset its aging workforce. AI art restoration shakes up the art world. Musk wants to block OpenAI's $97.4B bid documents. Anthropic will train on chat transcripts. TikTok cuts trust and safety staff for AI. I am an AI hater. Some thoughts on LLMs and software development. Are AI companies losing money on inference?

Last week's roundup

The B-word

OpenAI CEO Sam Altman has said that AI is in a bubble, comparing the current investor frenzy to the dot-com boom of the '90s, while criticizing "irrational" valuations for early-stage AI startups.

The big picture:

• Some of the latest bubble fears have been triggered by the highly anticipated GPT-5 making less of a splash than anticipated. Though the model performs better across several categories, OpenAI has admitted it was a less-than-stellar rollout.

• As a result, we're seeing renewed questions over whether we're hitting the limits of AI's scaling laws - whether simply adding more computing power and data would lead to artificial general intelligence.

• In addition, Meta has frozen all AI hiring after a months-long spending spree that saw compensation packages reach nine figures, suggesting even Zuckerberg is feeling pressure from investors concerned about spiraling costs.

• But the freeze could be related to internal disarray - Meta is also conducting its fourth AI reorganization in six months, with its AI staff now splitting into four specialized teams (AI research, "superintelligence" development, products, and infrastructure).

Subscribe now

Elsewhere in the FAANG free-for-all:

• Microsoft and the NFL extended their partnership to bring real-time game data and analysis to coaches and players using Copilot on Surface tablets and Azure AI.

• Meta rolled out AI-powered voice translations for creators on Facebook and Instagram globally, supporting English-Spanish translations with optional lip syncing.

• Meta may debut "Hypernova" smart glasses with a display for ~$800 in September, while Apple delays releasing immersive video shot around Vision Pro's debut.

• Microsoft added a COPILOT function to Excel for Beta Channel users, letting them use text prompts to categorize data, summarize feedback, and brainstorm ideas.

• And Google's Jigsaw and pollster Scott Rasmussen plan to use AI to survey Americans on political views to find common ground ahead of the US' 250th anniversary.

I’m afraid I can’t do that Dave

Anthropic has released a few new features to prevent Claude conversations from going off the rails - starting with the ability to end conversations after persistently harmful user interactions.

Why it matters:

• The ability for Claude to end its own conversations represents one of the first concrete implementations of "AI welfare" protections, even as Anthropic remains uncertain about Claude's potential moral status.

• Testing revealed Claude showed consistent aversion to harmful content, apparent distress with abusive users, and a preference to exit such interactions when given the option, suggesting some form of behavioral patterns around self-preservation.

• Overall, the company continues to add guardrails and safety features (in keeping up with its reputation as the "safe" AGI lab). It also partnered with the US National Nuclear Security Administration to build a classifier that blocks AI conversations about dangerous nuclear technology, the first major public-private effort to address nuclear proliferation risks.

Elsewhere in AI anxiety:

• Mustafa Suleyman warns that seemingly conscious AI is coming but should be avoided, arguing that developers must build AI for people rather than as digital persons.

• An MIT report reveals that 95% of GenAI pilots at companies have little to no financial impact due to learning gaps between tools and organizations.

• Grok's website accidentally exposes the underlying prompts for its AI personas, including one for a "crazy conspiracist" that frequents 4chan.

• Cybercriminals, spies, researchers, and corporate defenders are increasingly using AI to find software vulnerabilities in an escalating cat-and-mouse game.

• Texas Attorney General Ken Paxton is investigating Meta and Character.AI for allegedly misleading children by marketing chatbots as mental health tools.

• And AI-powered stuffed animals like Grem, Grok, and Gabbo are being promoted as screen time alternatives for children as young as 3.

Elsewhere in Anthropic OpenAI:

• OpenAI CFO Sarah Friar says the startup hit its first $1B revenue month in July and that "the biggest thing we face is being constantly under compute."

• OpenAI launches ChatGPT Go, a new India-only plan priced at Rs. 399, or ~$4.57, per month, its most affordable yet, to grow users in its second-largest market.

• A look at OpenAI's campaign over the past year to turn its for-profit arm into a public benefit corporation.

• Current and former OpenAI employees plan to sell ~$6B in stock to Thrive Capital, SoftBank, and others in a secondary sale that values OpenAI at ~$500B.

• And an interview with Sam Altman on spending billions on AI infrastructure, wanting Chrome if it's really for sale, AI models saturating the chat use case, and more.

Hybrid DeepSeek

DeepSeek released V3.1, an updated version of its V3 from earlier this year that's purportedly better than R1, the company's flagship reasoning model.

What to watch:

• The model appears to be a hybrid reasoning model (like GPT-5) that switches between reasoning and non-reasoning modes. It remains to be seen whether this type of model will become the industry standard moving forward.

• It's still early days, but the benchmarks don't appear to be earth-shattering - the model is in the same ballpark as OpenAI's gpt-oss model, and well behind SOTA models like GPT-5 and Grok 4.

• But with rivals like Alibaba's Qwen gaining traction and the delayed successor to R1 still pending, the Chinese AI landscape is becoming increasingly competitive, keeping pressure on both domestic players and Silicon Valley incumbents.

Elsewhere in frontier models:

• Nvidia debuts Nemotron-Nano-9B-v2, a hybrid Mamba-Transformer model that achieves comparable or better accuracies than Qwen3-8B on reasoning benchmarks.

• Alibaba releases Qwen-Image-Edit, the image editing version of Qwen-Image that supports modifying Chinese and English text while preserving font, size, and style.

• And NASA and IBM released Surya, an open-source machine learning model trained on over a decade of NASA solar data to predict solar flares and solar winds.

Made By Google

Google held its "Made By Google" event this week (hosted by, weirdly enough, Jimmy Fallon), and even though it was primarily about the Pixel 10 and the Pixel Buds 2, there were also several AI announcements during and around the event.

All the announcements I could find:

• Gemini for Home, a Google Home voice assistant with Gemini Live, available in early access in October and eventually replacing Google Assistant.

• A new Gemini-powered personal health coach for Fitbit, rolling out as a preview in October for Fitbit Premium users in the redesigned Fitbit app.

• AI Mode is expanding to 180 countries and territories in English, following the US, the UK, and India, with plans to add more languages and regions soon.

• New AI features for Pixel include Magic Cue for contextual suggestions across apps, Voice Translate for calls, and Visual Overlays for the camera.

• The Tensor G5, a chip made on TSMC's latest 3nm node with significant performance gains and support for the newest Gemini Nano model.

• Gemini for Government will be provided to US federal agencies at $0.47 per agency through 2026 (OpenAI and Anthropic are offering their products for $1/year).

• Google says the median Gemini app text prompt consumes 0.24Wh of energy, about the same as running a microwave for a second, uses 0.26mL of water, and emits 0.03g of CO2 equivalent.

• And the company reports users created 100M videos using its AI filmmaking tool Flow since its May launch, which leverages Veo 3 and focuses on maintaining visual consistency.

Things happen

Oracle on hook for tens of billions in data centers. AI Film Festival backed by Runway, screened by Imax. China's H20 curbs triggered by "best stuff" remark. Grammarly debuts AI agents. 71% fear AI will permanently displace workers. Claude still produces better code than GPT-5. Otter.ai sued for "deceptively" recording conversations. Anthropic raising $10B. AI designs novel antibiotics for drug-resistant bacteria. Stability AI goes Hollywood. Nvidia developing new China chip. Replacing junior staff with AI is "dumbest thing ever". Z.ai launches general-purpose AI agent app. Wales wants Wikipedia to use AI; editors revolt. Open-weight LLMs show inconsistent performance across hosts. AI assistant usage mirrors 2018. SoftBank buys Foxconn's Ohio plant for Stargate. AI doesn't lighten burden of mastery. Hawley to investigate Meta over AI chats with kids. GPT-5 suggests scaling LLMs hits limits. Executives mandate AI but haven't adopted it themselves. Baidu trades at ~9.7 P/E, lowest among profitable Hang Seng Tech companies. Meta's AI app hallucinates. Big Tech's reverse acquihires hollow out startups. The unbearable slowness of AI coding. Anthropic adds Claude Code to Enterprise. AI slop about DC creates disinformation mess. 87% of game developers use AI agents. Building agentic AI systems: best practices. 95% see zero return on $30B AI spend. Virtual salespeople outselling humans in China. AI books random meetings at games event. Model intelligence no longer constrains automation. Medium's evolving AI policy. AI is mass delusion.

Last week's roundup