Anthropic gives six METR evaluators employee-like access to its AI safety work Anthropic announced on October 15, updated October 16, that it has given six employees of nonprofit evaluator Model Evaluation and Threat Research (METR) employee-like access to company hardware, office space and internal systems used by its risk-assessment teams, with the right to publish assessments of Anthropic's risk levels, incidents and practices without Anthropic's approval. Anthropic retains the right to require redactions in predefined categories including customer data, attorney-client privileged information and model architectures, and has created three evaluator tiers in which first-tier partners need Anthropic's approval to publish while higher-tier evaluators get broader access and publication rights. The public announcement does not specify the METR agreement's funding, duration, termination protections or first-publication date, and does not give METR power to halt a model release or direct engineering work. Anthropic gives six METR evaluators employee-like access to its AI safety work - Anthropic says six METR employees have company hardware, office space and access to internal systems broadly comparable to those used by its risk-assessment teams.