Anthropic safety report highlights misuse of Claude for bioweapons and surveillance

Anthropic has released a comprehensive safety report detailing how threat actors have attempted to exploit its generative AI, Claude, for malicious purposes. The company identified seven primary harm areas, including cyber operations, influence campaigns, and biological misuse. Notably, the report documents instances where users attempted to leverage the AI to develop bioweapons, such as grant applications for researching dangerous viruses like chikungunya and orthopoxvirus. Furthermore, the findings reveal that state-sponsored actors have utilized Claude to engineer autonomous kamikaze drone swarms and develop electronic warfare modules. The report also highlights the use of the platform for mass surveillance, specifically in tracking dissidents and journalists. Anthropic stated that it has banned the accounts involved in these activities and is integrating these findings into its ongoing safety and monitoring protocols to prevent future exploitation of its technology.