Anthropic taps independent non-profits to evaluate artificial intelligence system risks

Artificial intelligence lab Anthropic announced on Thursday, October 15, 2026, that it is working with non-profit organisations METR and SecureBio to help assess risks and safety practices, CNBC reported. Tekedia and BigGo Finance both report that the company is giving independent evaluators access to evaluate its systems. CNBC added that Anthropic is also in early-stage discussions with other organisations, including Hugging Face. [1][2][4][6]

The partnerships give outside evaluators direct access to Anthropic systems. Tech-insider.org reported that Anthropic is giving six METR staff members access to its artificial intelligence system. NDTV reported that the access provided to METR is at an employee level, while BigGo Finance described the access given to the independent evaluator as unprecedented. [3][5][6]

The initiative follows a proposal published last month by Anthropic chief executive officer and co-founder Dario Amodei to slow down the pace of model development, according to CNBC. CNBC reported that Amodei said independent evaluators would help verify that Anthropic is adhering to its safety practices. Anthropic uses Constitutional AI, a technique introduced in 2022 that trains models using artificial intelligence feedback guided by natural-language principles. [1][7]

The artificial intelligence firm operates under a governance model that includes an independent Long-Term Benefit Trust, which is empowered to elect directors and protect its public benefit purpose against short-term pressures. Anthropic is working to embed third-party evaluators as part of its ongoing safety commitments, CNBC reported. [1][7]