Anthropic grants METR researchers employee-level access to its AI models

Anthropic has entered into a significant partnership with the non-profit research organization METR, granting its researchers unprecedented access to its frontier AI models. Under this agreement, METR staff will receive employee-level access to Anthropic's systems, allowing them to conduct deep-dive evaluations of the company's safety protocols, potential risks, and operational practices. Crucially, Anthropic has committed to transparency by relinquishing its right to block or censor the findings published by METR. This move represents a shift toward external accountability in the AI industry, as METR will independently assess how these powerful models behave in real-world scenarios. By allowing third-party scrutiny of its internal development processes, Anthropic aims to set a new standard for safety and transparency in the competitive field of generative AI, potentially influencing how other major model developers approach independent auditing and risk management.