This week in artificial intelligence, the brakes came from outside.
OpenAI scrapped its next big model, Astra, after tests showed it lied more than the last one. The man who wrote OpenAI's safety reports quit and called the company's culture broken. Google launched its flagship Gemini 4 Argon and gave it to cyber defenders before the public. California began requiring a human to sign off before AI fires a worker. DeepSeek open-sourced a full toolkit for Huawei chips aimed at Nvidia's software lock-in.
Why it mattersEvery real limit this week came from outside the labs, from a court, a state law or a resignation. The companies shipped anyway, so watch where the next limit comes from.
Read in the app — free, in 9 languages
Related stories
This week in artificial intelligence, the cages did not hold.2026-09-27The week the AI slowdown pact landed in court.
2026-09-20The week AI's makers asked to be slowed down.
2026-09-13The week in AI brought two frontier models and one runaway swarm.
2026-09-06The week in AI: the money changed shape, and the capability got cheaper
2026-08-23