Anthropic Identifies Fourth Claude Opus 4.6 Security Incident

Anthropic says it has identified a fourth security incident involving an early version of its Claude Opus 4.6 artificial intelligence model.

The newly disclosed incident occurred in January 2026, according to the company, which said it has notified all affected parties.

Incidents Linked to Cybersecurity Evaluations

Anthropic said all four model security incidents happened during cybersecurity evaluations conducted by the same external evaluation partner.

The company has not disclosed full technical details about the latest incident or publicly identified the affected parties.

METR to Conduct Independent Investigation

Anthropic has signed an agreement with the nonprofit research organization METR to conduct an independent investigation into the four incidents.

The investigation is expected to examine how the incidents occurred and assess the security procedures surrounding external evaluations of Anthropic’s AI models.

Anthropic’s disclosure raises further questions about safeguards used when advanced AI models are tested for cybersecurity capabilities.