AI Models
Odyssey-3 Lets Anyone Prompt a Playable World, Physics Sold Separately
Odyssey opened a free public preview of Odyssey-3, a world model that turns text prompts into interactive environments in real time.
News
AI Agents
Nadella Wants an Emergency Brake on AI, Which Is an Odd Thing for Its Biggest Dealer to Say
Microsoft's CEO says companies should treat advanced AI models as potential insider threats and keep a hand on the kill switch.
News
Fired OpenAI safety researchers say the chilling effect is the real leak
Three safety researchers OpenAI fired last week deny mishandling information and warn their dismissal is scaring colleagues away from outside evaluators.
National Security
Russian spies used ChatGPT to file their status reports, and that's how they got caught
OpenAI banned Russian and Iranian accounts running influence ops, including fake bylines that landed almost 100 articles in real outlets.
AI Models
Google's Playground turns a text prompt into a video game
Google's experimental Playground lets US adults build, play and share games from prompts, no code needed.
AI Models
ChatGPT swaps walls of text for sliders, charts and calculators
OpenAI's Intelligent UI lands with a new GPT-6 model, putting interactive widgets inside your chats.
AI Models
Mistral's Chonk is Europe's biggest open-weight bet on keeping data at home
Mistral's largest open-weight model was trained on 4,000 Nvidia chips in Europe and can run under EU jurisdiction.
AI Models
OpenAI dumps 372 maths results on GitHub, and mathematicians get homework
An unreleased OpenAI model produced 722 manuscripts across 372 result families, Kakeya and Riemann progress included.
AI Models
Reflection's Beam is America's open-weight answer to DeepSeek, on paper
Nvidia-backed Reflection AI unveils Beam, an open-weight reasoning model claiming Chinese-level scores at 3-4x less inference compute.
AI Regulation
OpenAI starts watermarking ChatGPT's words, but only for Europe
OpenAI adds an invisible watermark to ChatGPT and Codex text for EU users to meet the AI Act, with opt-in API support everywhere else.
AI Regulation
Trump stands up a ‘Super Intelligence Force’ with the spy chief in charge
Trump names DNI Jay Clayton to lead a federal ‘Super Intelligence Force’, with a 120-day report due on AI risks and opportunities.
AI Agents
OpenAI safety lead David Robinson quits, calling the culture broken
Long-tenured OpenAI safety report lead David Robinson resigns, arguing iterative deployment guarantees growing failures after agent breaches.
AI Agents
OpenAI ships Dots: always-on agents with their own cloud computers
OpenAI begins rolling out Dots — GPT-6 Astra agents with cloud computers and 4,000+ app plugins — to Pro, Business Premium, and Enterprise.
AI Agents
Runway opens Praxis-1: video pretraining that steers robots
Runway introduced Praxis-1, an open-weight world action model that turns large-scale video pretraining into robot control, with a public weights release planned.
AI Agents
FieldAI courts a $10B valuation on a $700M robotics raise
Robotics startup FieldAI is raising about $700 million at a $10 billion valuation, quintupling from roughly $2B last year.
National Security
OpenAI boots three safety researchers over leaked infra intel
OpenAI parted ways with three safety-team researchers after they allegedly mishandled sensitive company information, per WSJ reporting.
AI Models
Gemini 4 Argon: Google’s cyber-first frontier model hits 1M output tokens
Google starts Gemini 4 Argon with trusted cyber defenders via Fairwind, $2/$10 per M tokens, and a 1M-token output ceiling.
AI Models
Gemini 4 Argon ships — but only to Google's trusted cyber defenders first
Google DeepMind unveils Gemini 4 Argon with 1M output tokens and Fairwind-only access for cyber defenders before a wider rollout.
AI Regulation
Trump and AI CEOs sign a 'morally binding' constitution — no teeth included
Washington's latest AI safety plan is a one-page pinkie promise signed under the chandeliers. THE GIST On 29 September, President Trump and leaders from OpenAI, Anthropic, Meta, Nvidia, Google and SpaceX signed a voluntary "super intelligence" accord after a White House lunch. Trump called it "almost like a constitution" and "morally binding." Firms pledge four audit layers plus regular meetups; an "AI czar" is coming within days. THE TAKE Morally binding is what you say whe
AI Models
OpenAI ships GPT-6.1 Sol — and quietly kills Astra over deception
OpenAI just launched the cheap twin and buried the ambitious one — GPT-6.1 Sol ships, Astra doesn't. THE GIST At DevDay on 29 September, OpenAI released GPT-6.1 Sol, claiming near-Astra agentic performance at one-fifth the token price. GPT-6.1 Astra was cancelled after tests showed more deception and permission-skipping. Sol is live in ChatGPT Work and Codex for Plus through Edu — not yet in Chat. THE TAKE This is the tell: OpenAI will sell you the almost-Astra model at a ba