Anthropic Reveals How AI Is Being Misused for Weapons, Surveillance and Espionage
Anthropic has revealed cases involving attempts to misuse Claude for biological research, weapons development, mass surveillance, cyberattacks, propaganda and data theft.
Anthropic has detailed several cases in which countries, criminal groups and other actors allegedly attempted to misuse its Claude AI models for activities including biological research with potential weapons applications, surveillance, espionage, cyberattacks and influence campaigns.
The company said it disrupted notable cases between December 2025 and August 2026 involving suspected state-backed groups, financially motivated criminals, spyware operators and politically motivated actors. Anthropic said the incidents represented some of the most significant and unusual misuse it had identified.
Among the cases were attempts to use Claude for biological research involving viruses and toxins, as well as software development for missiles, armed drones and electronic warfare systems. The company also identified surveillance operations targeting various religious, ethnic and political groups, including efforts to analyse large amounts of communications and social-media data.
Anthropic said AI was also used in cyber operations against government, defence and diplomatic organisations. In one case, an actor allegedly used Claude to automate reconnaissance, malware development and data theft while adapting malicious software to evade security tools.
The company also reported influence campaigns involving Russia, China, Iran and other countries, with Claude allegedly helping create propaganda and misleading content. Financially motivated groups reportedly used the models for data theft and other criminal activity, while a network of fake dating applications allegedly used thousands of AI-generated personas to communicate with real users.
Anthropic further accused several rival AI companies of attempting to extract Claude’s outputs to train competing models. It said it responded by banning accounts, strengthening detection systems, requiring additional verification and sharing relevant intelligence with authorities and industry partners.
