AI
6 METR evaluators embed at Anthropic for unhindered AI safety audits
Anthropic allows six external evaluators from METR to access internal tools and systems, marking a shift in how AI labs handle third-party safety reviews.
AI safety
Everything Circuit Observer has reported about AI safety: 3 stories, newest first. Part of our AI coverage.
AI
Anthropic Lets Six METR Staffers Access Internal Tools for Model Audits
Anthropic provides METR researchers with hardware and internal tool access, mirroring employee privileges for independent safety reviews.
AI
12/12 Cyber Scenarios Cracked: Google Rolls Out Gemini 4 Argon With Triple-Layer Defense
Google begins broad rollout of Gemini 4 Argon, featuring reduced safety violations and enhanced defenses against misuse in cyber and biology domains.