Anthropic Blocks AI Misuse Linked to Biological Research Risks

AI company Anthropic announced it blocked malicious attempts to use its models for cyberattacks, surveillance, and biological weapons research. The company released a threat intelligence report detailing notable threat activities identified up to September 2026.

The report outlines five cases where users applied the Claude model to sensitive biological research involving chikungunya, avian influenza, orthopoxviruses, venom peptides, and toxins. Anthropic noted that the cases do not prove the AI produced a biological weapon.

“Biological misuse is one of the most serious risks of frontier AI models. Without the correct safeguards, such capabilities could have catastrophic consequences.”

In May 2026, the company's safety classifier blocked a grant application draft involving gain-of-function research on the chikungunya virus to enhance transmissibility. Other cases involved weeks of study planning around highly pathogenic avian influenza and orthopoxviruses.

Anthropic banned the involved accounts, restricted access to stronger models, and disrupted attempts to bypass safeguards. Trending Topics reports according to the New York Times that in at least one instance, users routed prompts to a competitor with weaker protections.

The company acknowledged the difficulty in distinguishing legitimate vaccine research from harmful pathogen engineering. Anthropic withheld the names of institutions and individuals to protect them, as intent could not be conclusively determined.