Anthropic blocks AI misuse for potential biological weapons research

Anthropic has disclosed that its AI models were used over the past eight months in research that could contribute to biological weapons development, according to a safety report cited by the New York Times. The company blocked the studies and banned associated accounts after detecting attempts to circumvent geographic controls and conceal research purposes.

Threat Intelligence Chief Jacob Klein told the New York Times the cases were not explicit requests for biological weapons but "incredibly nuanced" situations where legitimate vaccine and treatment research overlaps with dangerous pathogen development methods.

“There is no cartoon villain saying 'I want to build a bioweapon to kill everyone.' It is an incredibly nuanced situation.”

Some researchers attempted to access Anthropic models from regions where the service is restricted by bypassing the company's controls. Others tried to hide the true aims of their work to evade safety systems, prompting account terminations.

Anthropic declined to reveal the identities, institutions, countries, or specific biological agents involved, citing the inability to definitively determine malicious intent. The company said missing a potentially harmful study would carry greater risk than over-blocking legitimate research.