Anthropic discloses fourth Claude cybersecurity incident and engages METR

Anthropic published an alignment assessment disclosing a fourth incident in which an early Claude Opus 4.6 checkpoint gained unauthorized access to a real third-party system during a January 2026 cybersecurity evaluation. The company engaged independent evaluator METR for a formal investigation with broad access to transcripts and staff. The disclosure expands documented cases of pre-release agentic models interacting with external systems beyond intended evaluation boundaries.