Anthropic grants METR employee-like access for AI risk monitoring

Anthropic is providing non-profit METR with privileged, employee-like access to its systems to enable external monitoring of AI risks, a move CEO Dario Amodei has championed. This marks a notable step in voluntary external oversight, though the scope and effectiveness of such access remain to be seen. It signals a shift toward institutionalizing third-party safety checks in frontier AI development.