Anthropic is inviting independent evaluators inside the company permanently. They will have the same access as internal risk teams and the right to publish findings without the lab's editorial control.
It's part of the company's plan to "pace" AI development. Full story in
@FortuneMagazine