Accenture

Six METR staff now work inside Anthropic on company laptops, and Anthropic cannot veto what they publish

The arrangement gives an outside nonprofit standing access to the inside of a frontier lab, modelled loosely on the resident examiners that banking regulators place in banks.

Anthropic will send open-source maintainers vulnerability reports no human has read, and expects about nine in ten to be real

The scanner's reports go out model-generated and unreviewed. Days earlier, Google stopped taking AI-written bug reports for its own open-source projects.

Anthropic will spend $100m training engineers who work for its customers, not for Anthropic

The structure is copied from medical residency: an exam, then twelve weeks leading a real deployment inside your own employer, then a second assessment.

Anthropic and Accenture will each put $1bn into putting outside evaluators inside Anthropic

The evaluators get access comparable to an employee's. The choice of a consultancy over a safety non-profit is what the field is arguing about.