Anthropic Disrupts Malicious Attempts to Weaponize Its AI Models

Anthropic has thwarted multiple attempts by unidentified individuals to exploit its AI systems for cyberattacks, surveillance, and biological weapons research, the company reported Thursday.

Anthropic details blocked bioweapons-related requests. In one case its systems prevented a user from obtaining Claude's help drafting a grant proposal. The planned project centered on gain-of-function research on the chikungunya virus targeting its transmissibility and immune-evasion traits, according to the report.

Chikungunya is a mosquito-borne virus that causes debilitating symptoms such as severe pain and fever. The proposal sought funding to amplify mutations that would progressively increase the virus's harmfulness. While such work could "certainly" be used to develop better vaccines and treatments, Anthropic said, "it could also be used to make the pathogen more dangerous."

The company reports broader malicious AI misuse. As AI models grow more powerful, sophisticated cyberattacks no longer demand advanced expertise and even solitary actors can now generate threats impossible a year earlier, the company warned. It has introduced tougher filters in its newest releases to limit biological inquiries that might support weapon development.

"The cases we share here aren't typical misuse, but rather examples of the most notable and novel threat activity we've identified to date," Anthropic said in its third misuse report since March 2025. The document reproduces excerpts of the harmful prompts and code it detected and calls on governments and rival AI labs to help stop comparable exploitation.

Anthropic publishes its third misuse report since March 2025. The firm released the findings because it feels an obligation to reveal dangerous misuse of its technology. "As models become increasingly capable, their risks will increase, unless AI developers and society's defenders act to make them safer," it added.

New safeguards target biological weapon risks. Stronger protections were built into the latest models specifically to block biological research that could assist in creating weapons. This marks the company's third public disclosure on the topic since March 2025.