AI·September 11, 2026·2 min read

Anthropic blocks AI use that could aid biological weapons

Claude misuse cases expose growing biosecurity risk

Anthropic says it has disrupted attempts to use its Claude artificial intelligence models for research that could contribute to biological weapons development, highlighting a growing security challenge as advanced AI systems become more capable in biology and other sensitive fields.

The company detailed five case studies involving users whose work could support biological weapons development. One case involved a request for help writing a grant application for gain-of-function research involving the chikungunya virus, with proposed work focused on transmissibility and immune evasion. Another involved research on highly pathogenic avian influenza and how the virus adapts to mammals. Anthropic also identified an account preparing a grant application for orthopoxvirus research at a state-associated infectious disease laboratory.

Anthropic said it could not always determine whether the researchers had legitimate scientific goals or malicious intentions. That distinction is particularly difficult in biology because research that can help scientists understand viruses, develop vaccines or prevent disease can sometimes use techniques that could also make pathogens more dangerous. The company said it blocked or terminated accounts when activity crossed its safety thresholds rather than waiting to establish a user's ultimate intent.

The cases were part of a broader threat-intelligence report covering misuse detected and disrupted between December 2025 and August 2026. Anthropic reported attempts to use its models for cyberattacks, surveillance, propaganda, fraud and conventional-weapons work in addition to sensitive biological research. The company says newer frontier models require stronger safeguards because their growing scientific capabilities could provide more meaningful assistance in dangerous areas than earlier systems.

Anthropic has deployed real-time classifiers and other restrictions designed to detect dangerous biological requests while still allowing legitimate medical, educational and scientific uses. The challenge is maintaining access to potentially valuable AI-assisted research without giving malicious actors tools that could substantially lower the expertise needed for harmful work.

Community discussion
"How should AI companies balance access to advanced scientific capabilities with the risk that the same technology could be used for dangerous biological research?"