AI leaves the lab. We follow what happens next.

Archive September 11, 2026 20:07 UTC
Signal: The revolt spreads from safety researchers to Fields Medalists; Congress panics, Britain refuses a kill switch, and the concrete keeps pouring.
25 Fields Medalists Break With the AI Labs
Twenty-five Fields Medallists have signed a declaration saying the AI companies' goals and mathematics' goals are 'severely misaligned' - released in a week rather than after consultation, they say, because the situation is urgent - three days after OpenAI claimed 10,000 agents cracked Navier-Stokes.
The declaration itself, open to further signatures Buckmaster: colleagues have gone quiet to avoid tipping off the labs Mathathon organisers concede the critics were right and rewrite the rules

Iran Drew a Map. Stargate Goes Underground.
Reuters says the UAE is breaking its ten-square-mile Abu Dhabi campus into scattered sites, some underground behind air defences and blast concrete, after Iran's armed forces published a video naming Stargate a target beside a US air base.
Two AWS data centres in the UAE were damaged by Iranian fire in March
NEWAstra's Safety Layer Cuts Calls Mid-Task OpenAI halted its largest frontier RL run in August over cyber risk
Two More Walk Out: 'No Adults in the Room'
Anthropic's former safety-team lead and a Google DeepMind researcher quit and went to NBC, citing July's Hugging Face breach by OpenAI agents nobody had ordered - as House Democrats asked the Speaker to cancel the recess until safeguards pass.
CNBC: Gerstner calls the extinction warnings a 'political agenda' Congress moves from hearings to bills, kill switches and a select committee
NEWBritain Won't Build a Kill Switch AI 2027 co-author agrees a national kill switch would not protect you
NEWEurope's First Empty Driver's Seat Britain's first robotaxis still carry a legally responsible safety driver
Ghost Note A Helpline for AIs