Anthropic has granted a team from METR internal access to tools, workspaces, and information. This collaboration allows independent evaluation of potential risks associated with advanced artificial intelligence models.