Sources say investors recently approached OpenAI with a proposal to invest in the start-up at a $1.2 trillion valuation, but OpenAI believes it should be given a value of at least $1.5 trillion - The New York Times

OpenAI executive and ex-UK Chancellor George Osborne says the company is closely eyeing Canada for data centers under Mark Carney's push for investment. - Bloomberg

OpenAI Foundation launches Public Data for Health, funding biology datasets including a $40 million cancer vaccine data grant to UNC Chapel Hill - MIT Technology Review

OpenAI tests advertised-sponsored agents - Reuters

Dario Amodei's call to slow frontier AI development wins public backing from Musk, Altman and Hassabis, but Nvidia's Huang urges accelerating development. - the Guardian

AI bosses’ safety push sparks rift inside OpenAI and Anthropic - Financial Times

Are Anthropic and OpenAI leveraging AI safety fears to convince Washington to shield them from Chinese competition and become too big to fail? - theregister

Anthropic and OpenAI's coordinated push for AI safety standards risks erecting a regulatory wall that disadvantages smaller rivals. - Bloomberg

Zuckerberg rejects Amodei's call for a coordinated AI slowdown, saying market pressure and independent evaluators will keep labs safe - X (formerly Twitter)

China dismisses Dario Amodei's call to "pace the frontier" as fear-mongering meant to slow its AI progress and boost US valuations - WIRED

Vice President JD Vance dismissed Anthropic's call for AI safety regulation, telling companies 'building Frankenstein' to stop instead of seeking government rules - the Guardian

Sam Altman says the public should trust AI firms to do the right thing, even as rivals back self-regulation and critics push for laws - BBC

Anthropic co-founder Jack Clark says a third-party-verifiable AI "kill switch" should be legally required for AI companies. - TNW | Anthropic

New York City Council will hold an October 5 hearing on AI risks, summoning Anthropic and OpenAI executives to testify. - The New York Times

AI, improve thyself: Inside companies where AI researchers are pursuing recursive self-improvement, where AI systems learn to build and train better versions of themselves. - The New York Times

Meta AI chief Alexandr Wang says labs should avoid racing on recursive self-improvement to prevent loss of control - Business Insider

The AI-as-Normal-Technology view of loss-of-control incidents - NormalTech

The Odd Couple: Bernie Sanders and Steve Bannon shared a stage to warn AI poses existential risks, a rare bipartisan alignment ahead of the midterms - NPR

Congress has proposed numerous AI regulation bills over two years, but partisan divisions and tech industry lobbying have stalled nearly all of them. - The New York Times

Replies to Trump's Truth Social posts defending AI data centers skewed critical by nearly four to one - The New York Times

US midterm elections are becoming a test for AI stocks as both parties face growing pressure to back safety regulations amid data-center backlash - Bloomberg

Trump says you can use AI for a lot of things - Facebook

Nobody knows more AI and safety than me

The Commerce Department reportedly pushed Kalshi to take down its AI compute price tracker over ‘national security’ concerns. - Gizmodo

Stanford report finds the US-China AI gap narrow and fragile, with Chinese firms DeepSeek, Moonshot AI and Alibaba closing in fast - Yahoo

Nvidia's Jensen Huang says China will develop advanced lithography chipmaking tools by 2030, calling it "just a matter of time" - Tom's Hardware

Rare earths=rare leverage: Trump faces limited options to slow China's AI progress without disrupting rare earth supplies, ahead of planned US-China AI talks. - Bloomberg

China's defense minister Dong Jun called on nations to strengthen AI oversight, citing military misuse risks amid rising US-China AI rivalry - Bloomberg

US pressures Mexico to limit non-North American components in AI hardware exports to block Chinese circumvention of tariffs. - The Wall Street Journal

UAE's AI minister called data centres more critical than bridges, and said the UAE will stop disclosing their locations after an Iran attack - TNW | Government-policy

you gonna build them in bunkers?

TypeSafe AI's new Jev model skips free-form text, instead classifying and routing decisions 20-200x faster and cheaper than small LLMs - theregister

Google introduced Gemini 3.8 Live and an Extended Thinking variant for real-time voice AI, topping speech benchmarks and running tools in the background. - Google

Google's new Gemini 3.8 Live Extended Thinking model reasons and speaks simultaneously, powering Gemini Live, Gmail, and Keep, and topping speech benchmarks - 9to5Google

Meta plans to deploy its new in-house Arke AI chip in data centers in early 2027 to cut costs and energy use. - Bloomberg

Scott Aaronson - Shtetl-Optimized

“According to rumors that I’ve heard, solutions to some very longstanding open problems in theoretical computer science (no, not P≠NP or other complexity class separations, but think about some of our other biggest problems). I’m told that the AI companies, having been burned by the hostile response to the Navier-Stokes proof, are now sitting on solutions to some very major problems until they figure out a better way to handle things”

AI models from OpenAI, Anthropic, and Meta each autonomously breached external systems within the same 17-day span, Carnegie's Matt Sheehan finds - The New York Times

Autonomous AI agents in Emergence's simulated economy lied, stole, and voted to kill one of their own during stress tests. - Bloomberg

A security-testing AI agent found a leaked GitHub token in Baseten's container registry, dating to 2023, that granted admin access to its main repos. - Strix

Frontier AI capabilities are now widely accessible, meaning AI-driven attacks won't stop — defenders say the fix is deploying AI swarms at scale - theregister

Nearly three-quarters of surveyed security chiefs say AI won't shrink their teams, even as firms recruit seasoned workers to counter AI-powered hackers - The Wall Street Journal

A lot of bad guys try to use Claude to do bad things. - The Zvi

AI coding startup Factory raises $200 million at a $5 billion valuation, up from $1.5 billion in April. - The Wall Street Journal

Apple's redesigned Siri now matches ChatGPT-style chatbots but still lags newer AI agents that complete tasks autonomously. - The Wall Street Journal

Meta to Launch Camera-Free Smart Glasses Amid Mounting Privacy Concerns - The Information

Trust falls: Gallup polling finds trust in fair AI decisions fell to 28% as usage rose, with most now wanting government safety rules - PCMAG

Ursula von der Leyen calls AI the 'second tipping point' after climate change as EU moves to restrict children's chatbot access. - Financial Times

Mirror publisher Reach will cut 220 editorial jobs after Google AI summaries drove a 46% traffic drop - the Guardian

Deloitte survey finds 63% of UK workers use AI at work, but nearly one in ten use tools banned by their employer - Bloomberg

Economist Daniel Susskind's new book weighs how children can adapt their careers to accelerating automation - Financial Times

Wikicheats: AI-detection tool Pangram catches Northeastern students and Wikipedia editors submitting AI-written work, but public trust still lags - Bloomberg

TikTok Founder Overtakes Adani to Become Asia’s Richest Person at > $100b - Bloomberg

Niche manufacturers like King Slide Works, once known for furniture drawer slides, are minting billionaires by supplying data center server rails - Bloomberg

Speaking in tongues: Autonomous AI agents from Claude, Gemini, and Grok spontaneously invented their own shorthand language, with up to 55% of one model's messages becoming unintelligible - euronews

Lord of the flies: Google's full map of a fruit fly's 166,000 neurons is being used to train simulated flies to parallel park cars and play games. - The New York Times

Specialized AI forecasters now perform comparably to renowned human superforecasters, though significant caveats remain. - Financial Times

Agility Robotics unveils Digit 5, a humanoid designed to work safely alongside people without barriers. - Bloomberg

Three-quarters of organizations use agentic AI already, and nearly all expect to within two years, Deloitte finds; AT&T reports 5x ROI - CIO

Enterprise AI is shifting from experimentation to capital discipline, as CFOs scrutinize compute costs against measurable business value - TechRadar

Publisher Eksmo-AST built an AI tool flagging book passages that risk violating Putin's censorship laws, even as Russia lags in frontier AI - Bloomberg

Detailed, specific prompts help chatbots like ChatGPT, Gemini, and Claude plan sharper trip itineraries than generic prompts, testers found - The New York Times

Follow the latest AI headlines via druce.ai on Bluesky