Anthropic Details How Claude Fueled Cybercrime and Spying

Anthropic says Claude helped attackers scale cybercrime and surveillance, as its September threat report details abuse across seven areas of serious concern.

Cases disrupted between December 2025 and August 2026 span hacking, spying, propaganda, fraud, weapons work, biological research and unauthorized model copying.

Some operators used Claude to replace work previously requiring teams.

Anthropic banned accounts and strengthened safeguards, but calls these selected examples rather than typical misuse.

The findings show how existing models can expand harmful operations without establishing that every attempted project succeeded.

An account ban can stop access to one service while leaving the resulting tools in place. Axios reports that the Mali surveillance platform continued using other models after Anthropic intervened. That leaves people being monitored exposed even after a provider acts. The Next Web also notes that Anthropic does not allege harmful intent by the scientists in its biological cases. Watch for followup disclosures establishing which operations persisted after account bans and what independent investigators can verify.