Anthropic disrupted attempts to use its AI models for biological research with potential dual-use applications, including cases tied to Iran and studies on pathogens like chikungunya virus, according to a 154-page report and multiple outlets.
Anthropic has blocked efforts by researchers to use its AI models for biological research that could aid in developing biological weapons, according to a 154-page report and statements from the company. The firm identified five case studies involving researchers who used its Claude chatbot for advanced biological research with dual-use potential, including work on gain-of-function studies of the chikungunya virus and bird flu, the company said.
Dual-Use Research and Biological Weapon Risks
The report highlighted the “dual-use” nature of biological research, where tools and knowledge can serve both medical and malicious purposes. One case involved a researcher asking Claude to draft a grant proposal for gain-of-function research on the mosquito-borne chikungunya virus, which involves modifying the virus to make it more transmissible and dangerous. While such research can help develop vaccines, the company warned it could also be misused to create more harmful variants.
The same information that can be used to develop a biological weapon could also be used to develop, for example, a vaccine or a cure for a disease,
the report stated. Jacob Klein, the head of threat intelligence at Anthropic, described the situation as an incredibly nuanced situation,
noting that malicious intent is not always overt. You are not seeing someone in a comic book kind of way say, 'Hey, I want to build a biological weapon to kill everybody,'
he said.
Iran-Linked Cases and Surveillance Threats
Anthropic also reported disrupting an Iran-linked threat actor that used its AI models to collect and analyze public data for targeting recommendations against U.S. naval forces in the region. The company banned the actor’s account and shared threat intelligence with government authorities. The report noted that such actors exploit the plausible deniability
of dual-use biology research to mask malicious intent.
The firm’s findings align with broader concerns about AI misuse. In August 2026, a senior Anthropic safety researcher warned that AI has a greater than 10% chance of 'kill[ing] all humans' within the next decade,
a claim that sparked internal debate. Former researcher Jacob Coxon, who resigned, criticized the company for racing straight to self-improving superintelligence and gambling with our lives,
according to a source.
Global Surveillance and State-Sponsored Misuse
Anthropic’s report also detailed cases of AI misuse in surveillance and espionage, including efforts by actors in China and the Iran to exploit its models. The company claimed Chinese firms like Moonshot used its technology to train their own models by routing user queries through fake accounts. In Iran, researchers allegedly used Claude to develop tools for identifying dissidents via social media profiles.
These cases reflect a broader trend of state-sponsored AI misuse. The firm noted that Sophisticated threat actors are aware that we (and other AI providers) are attempting to detect dangerous uses of our models, and they use the dual-use nature of biology to maintain a kind of ‘plausible deniability’ about their research,
the report stated.
Ethical Dilemmas and Regulatory Calls
The incidents have intensified calls for stricter AI regulation. U.S. Senator Bernie Sanders has demanded a pause on advanced AI development, while President Donald Trump expressed concerns about falling behind in AI competition. if we don’t win AI, we’re going to be put in a very bad position,
he said.
Anthropic emphasized that it is not alleging malicious intent in the cases it blocked but stressed the need for safeguards. The company banned associated accounts, shared findings with government authorities, and strengthened its safety measures. Biological capabilities are dual use: they can be used for beneficial or harmful purposes, and it is often difficult to distinguish between them,
the report stated.
What Comes Next: AI Governance and Oversight
The incidents underscore the urgent need for global AI governance frameworks. As AI models become more powerful, the line between beneficial and harmful applications grows blurrier. Anthropic’s actions highlight the challenges of balancing innovation with security, particularly in an era where state and non-state actors alike seek to exploit emerging technologies.
The company’s report, published in 2026, comes amid growing scrutiny of AI’s societal impact. With the potential for both life-saving breakthroughs and catastrophic misuse, the conversation around AI ethics and oversight remains critical. As one source noted, The same information that can be used to develop a biological weapon could also be used to develop, for example, a vaccine or a cure for a disease.
