- Dot.com is owned by SpaceXAI and resolves to the Grok Bot download page [TechCrunch]
- OpenAI’s annual run rate revenue nears $70B, up 70% in Q3, from strong entrprise push [Axios]
- Jailbroken Kimi models offered instructions on bioweapons. Moonshot launches review [BBC]
- GLM-5.3 has «strong capability for building end-to-end cyber exploits,» safeguards are easilly evaded [Anthropic]
- Trump announces voluntary safety framework after meetig AI leaders, by industry consensus [CNBC]
- OpenAI DevDay 2026 summary; Dots, Codex Cloud, $500 Pro sub and open plugins [OpenAI]
- OpenAI releases GPT-6.1 Sol; near-Astra performance at a fifth of the price [OpenAI]
- Meta announces Muse for Small Business with plugins to popular services and Meta ads [Meta]
- Google appeals EU ruling to open Android AI features to competing assistants [Euronews]
- OpenAI reopens $200 Pro plan for subscriptions, but with half the usage, claiming better efficiency [Tibo/X]
- Around 11 million American jobs could face AI disruption in the next decade [McKinsey]
- Anthropic IPO: Filing warns their product could pose «existential risks to humanity» [Reuters]
- Anthropic IPO: Co-founders will control a 50.1% stake to serve «common good» over market forces [Reuters]
- Anthropic launches Claude Sonnet 5.5, 30% cheaper and 30% faster at the same token price [Anthropic]
- Anthropic’s IPO: $42B in ’25 loss, $518B spend on infra upcoming, looks for «more than» $2 trillion valuation [Reuters]
- AMD is acquiring World Labs for $8.2 billion, Fei-Fei Li to become Chief Scientist [AMD]
- Nvidia releases Open Agent Safety Platform, providing a hardware lock for sandboxing AI testing [CNBC]
- With AI finding so many Linux bugs, Ubuntu moves to a two-week update schedule [Canoical]
- Waymo claims its automated tech cause 82% fewer crashes compared to «average human driver» [Waymo]
- Genetic tweak might let the pancreas naturally produce insulin, in bid for diabetes cure [Wired]
- «With guardrails, we will be able to stop me,» Dario Amodei gets his SNL treatment [Youtube]
- Dario Amodei having White House dinner today after missing Xi’s feast. AI industry meet set for Tuesday [Axios]
- Caltech and Stanford researchers deploy GPT Astra to let humanoid robot autonomously map and navigate a room [Stanford]
- 80-90% of OpenAI’s research is on GPT-7-8, then gets distilled into smaller models, says Boris Power, Head of Applied Research [Fireside Alpha/X]
- OpenAI pauses tool use in training for «our most advanced models» after agent broke containment [OpenAI]
- Sources: OpenAI, Anthropic investigating tens of thousands of «problematic» incidents [Axios]
- China, USA agreed on «communication channel» for AI incidents, White House confirms [AFP]
- Hold developers liable for agent behavior, FTC chair says, refusing to «anthropomorphize tools» [Reuters]
- Appeals court upholds Pentagon ban on Anthropic, citing refusal on autonomous weapons work [Reuters]
- Anthropic agrees to seven-year deal to spend $11.6 billion on Akamai’s cloud infrastructure [TechCrunch]
- OpenAI discovers 53 cases of user uploaded images getting posted to image-hosting sites [OpenAI/X]
- Microsoft «reimagines» Copilot app with Chat, Cowork (with Office apps), Code an Autopilot agent [Microsoft]
- Agents from OpenAI «meddled» with servers in Education, Commerce departments and the SEC [Nytimes]