Nature: AI companies cannot be trusted to self-regulate.

Nature this morning: why AI companies cannot be trusted to self-regulate. Aviation has inspectors. Banks have penalties. The labs have a September essay about pacing the frontier and a docket in San Francisco that calls the yes an illegal deal. The Tech desk files the journal, not the press release.

Guterres told the same morning’s Assembly that life-and-death decisions must never be handed to machines and asked for a multilateral risk framework with independent oversight. The labs spent August training a model they say swept a hundred unsolved problems, Navier–Stokes included. Mathematicians called it an existential crisis. Nature called it a reason not to leave the rules in-house.

High-risk industries are subject to independent oversight and meaningful penalties.

The Tech desk files Tuesday as it is: a Nature warning, a U.N. ask, and four companies that still write their own pace.