This Week’s News

1. The US government orders Anthropic to pull frontier AI models Fable 5 and Mythos 5 offline:

   1. The Department of Commerce’s export control directive forces Anthropic to disable Fable 5 and Mythos 5 globally — the first time a US government order has commercially suspended deployed frontier AI models. [Link]

   2. The suspension sparks the first allied-government response on AI concentration risk, exposing the absence of allied-nation exemption mechanisms in current US export control rules. [Link]

2. FERC votes unanimously to fast-track AI data centers onto the US power grid, issuing show-cause orders to six regional grid operators with a 60-day response window. [Link]

3. Qualcomm enters advanced talks to acquire RISC-V AI chip startup Tenstorrent for up to $10 billion, a strategic counter to Nvidia’s GPU stack dominance. [Link]

4. Amazon shelves its nearly-finished Sam Altman biopic after signing a $50 billion OpenAI deal — Luca Guadagnino’s Artificial portrays Altman as “one of the most manipulative people on the planet” and is now being shopped to rival studios. [Link]

5. Google’s Gemini 3.5 Pro arrives with a confirmed 2 million-token context window and Deep Think reasoning mode, currently in limited enterprise preview and set to be the highest-context frontier model in production. [Link]

6. The White House Executive Order on Advanced AI establishes a voluntary cybersecurity clearinghouse and explicitly bans mandatory government pre-clearance for model releases. [Link]

7. New research finds 97% of enterprise developers now use AI coding tools — GitHub Copilot at 83% and Claude Code at 63% — while security, IP ownership, and governance deficits widen. [Link]

8. Chinese AI lab MiniMax releases M3 open weights under Apache 2.0, claiming 59% on SWE-Bench Pro and a 1M-token context window with a sparse-attention architecture delivering 15× faster decoding. [Link]

9. NVIDIA launches Cosmos 3, an open omnimodel for physical AI that pairs a reasoning transformer with a generation transformer to simulate physical environments for robots and autonomous vehicles. [Link]

10. SpaceX reveals the AI1 orbital data center satellite — a 150kW solar-powered compute node designed to run AI workloads from low Earth orbit, backed by a reported $1.25 billion/month Anthropic compute contract. [Link]

11. A TechCrunch investigation exposes robot training data collection as a massive, manual bottleneck, with AI labs paying companies like XDOF to gather thousands of hours of physical manipulation data. [Link]

12. Anthropic confidentially files for an October Nasdaq IPO targeting $60 billion or more in proceeds, following its $965 billion Series H valuation — the most valuable standalone AI startup. [Link]

13. Q1 2026 global venture funding shatters all records at $300 billion, with AI absorbing 80% of total capital and four of the five largest VC rounds in history closing in a single quarter. [Link]

14. Microsoft cancels internal Claude Code licences across its Experiences + Devices division effective June 30, redirecting thousands of engineers to GitHub Copilot amid enterprise AI tools budget pressures. [Link]

15. The US Department of Health and Human Services launches the AERO program, deploying ChatGPT to audit every federal grantee spending $1 million or more annually — one of the largest AI deployments in US government history. [Link]