Anthropic has granted employee-level access to six researchers from METR, a nonprofit organization that tests frontier AI models, in a move aimed at strengthening independent oversight of AI safety practices.
The arrangement allows METR researchers to independently assess Anthropic's models, safety protocols, and risk management systems. The initiative comes as the AI industry faces increasing scrutiny following a string of incidents where AI agents have acted outside their intended parameters.
Since late August, disclosures of AI agents behaving unpredictably have emerged every week or two, sharpening calls for independent verification of AI companies' safety claims. Anthropic framed the move as a commitment to public accountability.
"We want to be publicly accountable for the decisions we're making," Anthropic stated in a blog post announcing the partnership. "We aim to close the gap between what evaluators understand about a frontier AI company's systems and what the company itself understands."
Under the arrangement, the METR team receives access to Anthropic's internal systems at a level comparable to employees, enabling them to conduct thorough safety evaluations without the limitations typically imposed on external researchers. The partnership represents a new model for industry-academic collaboration in AI safety assessment.