Anthropic has disclosed that it blocked attempts to misuse its AI systems for biological weapons development, including research on chikungunya virus transmissibility and immune evasion.
The company identified multiple accounts attempting to use its AI models for gain-of-function research on infectious diseases, such as bird flu, with a focus on viruses’ adaptation to mammals and severe illness in mammals. One researcher outside the U.S. used Claude for bird flu research.
Over a 30-day period, Anthropic detected 35 distinct research efforts with potentially concerning biological misuse activity, though the intent to harm remained uncertain. The blocked efforts also included research on orthopoxviruses—such as smallpox and mpox—and venom toxins.