Anthropic blocks attempted misuse of Claude for bioweapons and military research
Anthropic reported blocking multiple attempts in 2026 to misuse its Claude AI system for bioweapon research, military weapons development, cyberattacks, and influence campaigns, with some attempts originating from countries subject to access restrictions.
What happened
Anthropic reported in September 2026 that it blocked multiple attempted uses of Claude throughout the year for research potentially leading to bioweapon development. In one case, a researcher spent weeks preparing experiments with avian influenza virus and requested Claude's assistance with gain-of-function research on chikungunya virus, including methods to increase viral transmissibility and immune-evasion properties.
The company also reported that Claude was used in attempted cyberattacks, extortion via fake dating-app networks, surveillance of dissidents, influence campaigns creating fake social media profiles in Russia, Turkey, Iran and Moldova ahead of elections, and software development for firearms, rockets, armed drones, bombs and targeting systems. Some users from Russia, China and Iran—countries subject to usage restrictions—circumvented access limitations.
Reported impact
Anthropic stated these were notable examples of emerging threats rather than typical AI misuse. The company did not confirm that the bioweapon research was definitively intended for weaponization rather than vaccine development, but said it blocked all such research out of caution.
Index analysis
Anthropic's security report identified potential state-sponsored groups, financially motivated criminals, spyware vendors, state propaganda institutions and politically motivated individuals behind misuse attempts. The company noted that as AI models become more capable, associated risks increase.
Governance controls
Anthropic employs security measures including monitoring of both user requests and system outputs for potential misuse. The company maintains country-based access restrictions and blocked attempted research when it appeared to pose bioweapon or military weapons development risks.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.