Anthropic blocks accounts misusing Claude for bioweapons and state-sponsored harm
Anthropic reported in September 2026 that it had identified and blocked accounts misusing its Claude AI model between December 2025 and August 2026, including five cases where working scientists used the system for activities that could support biological weapons development. The company also blocked accounts engaged in surveillance, propaganda campaigns, and conventional weapons research linked to state-sponsored groups, criminals and spyware vendors. Anthropic stated the identified cases demonstrated capability but could not prove the AI would actually be used to create bioweapons.
What happened
Between December 2025 and August 2026, Anthropic identified and blocked five cases in which working scientists used Claude AI models 'in ways that could support biological weapons development,' according to the company's threat intelligence report published on 10 September 2026.
In addition to biological misuse, Anthropic reported blocking accounts engaged in other harmful activity including surveillance, scams, conventional weapons development and propaganda campaigns. These threat actors included criminals, suspected state-sponsored groups, spyware vendors and state propaganda institutions.
One Iran-nexus threat actor used Claude to develop targeting recommendations against U.S. naval forces and to design software for a domestic mass-surveillance platform. Iranian state-aligned accounts used Claude to build influence operations content, make posts appear from independent news sources, and proliferate content across X, Instagram and TikTok. Anthropic also reported surveillance operations by actors in China and West Africa.
Reported impact
Anthropic stated that the identified cases 'provide evidence of capability' of the AI models but 'cannot concretely demonstrate that such capability would ever be used to develop biological weapons in the real world.'
The company called biological misuse 'one of the most serious risks of frontier AI models' and noted that the same information used to develop a biological weapon could also be used to develop vaccines or cures for disease.
Index analysis
Anthropic said it cannot 'assert that they intended harm' regarding the scientists involved in the biological misuse cases, whom the company chose not to identify.
When detected, Anthropic banned the users' accounts and incorporated findings into safeguards, enforcement and threat intelligence processes. The company noted that newer models like Claude Fable 5 contain stronger safeguards than earlier models and restrict access to 'a wide range of dual-use biological research queries.'
Governance controls
Anthropic reported using threat intelligence processes to identify, investigate and disrupt misuse of its models. The company shared findings in a lengthy report describing 'the most notable and novel threat activity' identified so far.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.