techcrunch.com
Anthropic is letting METR researchers look under the hood with employee-level access
Anthropic announced that six researchers from the nonprofit METR will receive employee‑level access to its internal systems, tools, and office space to conduct independent evaluations of the company’s AI models, safety protocols, and risk management. The partnership follows a series of high‑profile incidents involving AI agents acting beyond their instructions, intensifying demands for external oversight. Anthropic will not block METR’s publication of findings, though it may request redactions for sensitive data. The move aligns with a broader [...]