Anthropic is allowing independent researchers from the non-profit METR to access its internal systems for security audits. This initiative, proposed by CEO Dario Amodei, grants METR staff-level access to development environments and risk assessment tools. Anthropic cannot block the publication of findings, though it can request redaction of sensitive information like model architecture or personal data. AI
IMPACT This move could set a precedent for greater transparency and independent oversight in AI safety practices across the industry.
RANK_REASON Significant policy shift by a major AI lab to allow external security audits. [lever_c_demoted from significant: ic=1 ai=1.0]
Read on dev.to — Anthropic tag →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →