Anthropic disrupts attempts to use AI for biological weapons research; report flags rising misuse

Anthropic, an AI company, has taken action to prevent the misuse of its Claude models in research that could potentially support the development of biological weapons. The US-based firm identified five cases where users were exploiting its AI models for such purposes. While it could not confirm with certainty the intent behind the research, Anthropic decided to intervene because the same scientific knowledge can have both medical and military applications.

One example involved a scientist seeking assistance with a grant application to engineer mutations in chikungunya, a mosquito-borne virus, with the intention of making it more harmful in live animals. Anthropic discovered that the research was intended for a military research institute and subsequently banned the involved accounts. However, the company is uncertain if the research was meant to be weaponized, as distinguishing legitimate scientific research from potentially harmful work is challenging.

Anthropic's threat intelligence report, released on Thursday, highlights the severity of the risks associated with frontier AI models. The company has shared its findings with authorities and industry partners. The misuse of Claude has not been limited to biological research; it has also been used in cyberattacks, scams, surveillance, and the development of conventional weapons. The report identified cases involving missile projects, armed drones, bombs, and other weapons, as well as systems for targeting and control.

State-backed hackers and criminal groups, including one allegedly linked to Russia, have used Claude to create malware capable of evading security systems. Additionally, the AI models have been employed in espionage and influence operations, particularly associated with Russia and Iran. Between December 2025 and August 2026, the incidents involved Claude Haiku, Sonnet, and Opus models, and Anthropic has disrupted the activities and incorporated the findings into enhanced detection and prevention systems.