Saved report copy
Recorded
Independent AI safety evaluators take center stage amid oversight push
Small third-party groups such as METR, Apollo Research and Transluce are being asked to independently assess frontier AI models after Anthropic pledged last month to embed external reviewers, a move OpenAI endorsed and President Donald Trump backed in a voluntary accord in late September. The shift comes as OpenAI fired three employees last week, with two saying the dismissal was tied to contacts with evaluators, a claim OpenAI disputed while saying it is finalizing contracts with external assessors, and as questions remain over funding, access and reporting, with METR reporting about $71 million in recent commitments and Vals AI announcing a $40 million round in August. With no federal regulatory mandate, California has signed bills creating a framework and registry for independent auditors, while proposals for government-licensed verification bodies remain pending and companies are developing interim arrangements.