Anthropic said Thursday it blocked attempts to misuse its artificial intelligence models for cyberattacks, surveillance, political influence operations and biological research that could potentially contribute to the development of dangerous pathogens.
The AI company said increasingly capable models can lower the technical skills needed to carry out sophisticated cyberattacks, allowing individuals and smaller groups to create threats that previously required more expertise.
Anthropic detailed the findings in its third report on AI misuse since March 2025. The company said it identified and blocked the activities and shared information with government agencies and industry partners to help prevent similar abuse.
“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.
The report covers activity identified between December 2025 and August 2026 involving actors that Anthropic said included spyware vendors, politically motivated individuals and state-sponsored groups.
AI request linked to virus research
Anthropic said one user attempted to use its Claude model to help prepare a grant application for gain-of-function research involving chikungunya virus.
The proposed research sought to alter the virus to increase its transmissibility and ability to evade immune responses. Anthropic said the research could support the development of vaccines and treatments but could also potentially make the pathogen more dangerous.
The company said its systems blocked the request.
Anthropic said older models such as Claude Opus 4 and Claude Sonnet 4.5 had fewer safeguards for advanced biological research because they lacked the capabilities needed to meaningfully assist sophisticated users with dangerous research.
The company said newer models can assist with more complex scientific work, prompting it to strengthen safeguards that restrict access to a broader range of dual-use biological research requests.
Influence operations
Anthropic also identified groups that created hundreds of social media accounts designed to appear as ordinary users before posting material supporting the same political viewpoints.
The company said it identified nine such operations originating from Russia, Iran and Turkey, as well as countries across the Persian Gulf, South Asia, Africa and Europe.
Anthropic said AI platforms could potentially identify such operations while they remain in development, before the activity becomes visible on social media.
Report follows researcher’s resignation
The report came a day after Anthropic researcher Jacob Coxon announced his resignation over concerns about the pace and direction of AI development.
Coxon said Anthropic and OpenAI were “racing straight to self-improving superintelligence and gambling with our lives” and warned that some researchers believe AI could pose a threat to human life before the end of the decade.
Anthropic said it used the misuse cases identified in its latest report to strengthen its safeguards.
“We hope that the findings in this report will help other developers recognize similar patterns on their own platforms, give governments and civil society a clearer view of how emerging threats take shape, and strengthen collective defenses,” the company said.
