Anthropic released its most comprehensive threat intelligence report yet, outlining how the company identified and disrupted advanced efforts to misuse Claude for cyberattacks, covert influence operations, surveillance, biosecurity threats, and weapons risks.

Every operation highlighted in the report was intercepted before reaching its objective. Anthropic applied lessons from these incidents to harden its safeguards while sharing relevant threat data with peer institutions and authorities.

Though these high-severity incidents are rare, Anthropic published the data to shed light on emerging threat vectors, demonstrate where safety systems succeed, and pinpoint areas needing refinement. By sharing these findings, the company aims to help the broader tech ecosystem spot similar activity and provide the public with a clearer picture of emerging AI risks.

For more details, read the full report here or download the PDF version.