Anthropic cautions that the misuse of AI may facilitate the development of biological weapons

Anthropic states that potential misuse of AI encompasses efforts to aid in the development of biological weapons, engage in cyber operations, conduct surveillance, and perpetrate scams and fraud.

Anthropic has reported that it has recognized and thwarted efforts to misuse its artificial intelligence models in ways that could aid in the creation of biological weapons.

The company’s latest threat intelligence report, published Thursday, revealed instances of suspected misuse of its Claude AI models occurring between December 2025 and August 2026.

Anthropic reported that it has identified five instances in which individuals utilized its models in ways that could potentially aid in the development of biological weapons. The company characterized biological misuse as “one of the most serious risks of frontier AI models,” cautioning that without suitable safeguards, such capabilities could lead to catastrophic outcomes.

“The same information that can be utilized to create a biological weapon could also be employed to develop, for instance, a vaccine or a cure for a disease,” Anthropic stated.

Jacob Klein, Anthropic’s head of threat intelligence, conveyed to the New York Times that the matter was “an incredibly nuanced situation,” emphasizing that malicious intent is not always readily apparent.

“You’re not encountering someone who dramatically declares, ‘Hey, I want to create a biological weapon to wipe out everyone,’” he stated.

The report identified six instances in which Claude was reportedly utilized to create software for conventional weapons, including firearms, missiles, armed drones, bombs, and other munitions, along with systems employed for targeting and control.

Anthropic reported that its models have been misappropriated for various purposes, including cyber operations, surveillance, scams, fraud, and influence campaigns. Over the last eight months, cases have emerged involving counterfeit dating applications, scams related to hotel Wi-Fi, and surveillance systems aimed at identifying dissidents.

The company stated that Claude had been utilized by actors associated with a cyber espionage campaign based in Russia and an institution involved in Iranian propaganda. It also identified the hacking group ShinyHunters and laboratories based in China as being linked to the misuse of its technology.

According to the report, a hacking group that aligns with the activities of the Russia-based Midnight Blizzard is said to have utilized AI to create a system that can identify when its malware has been detected by security systems and modify the code to avoid detection.


Anthropic has also alleged that Chinese AI companies are trying to duplicate Claude’s capabilities.

The company stated that it had integrated the findings into its security processes “to enhance the prevention, detection, and disruption of these activities in the future” and had communicated relevant intelligence with authorities and industry partners.

The report emerges alongside rising apprehensions within the technology sector regarding the possible dangers posed by ever more potent AI systems.

Anthropic stated that none of the reported misuse cases were associated with its Claude Fable or powerful Mythos-class models, except for one instance involving distillation, a method where smaller AI models are trained using larger models.

The company's warning comes in response to distinct concerns expressed by AI researchers and industry leaders regarding the speed of development.

OpenAI chief scientist Jakub Pachocki has recently advocated for “voluntary slowdowns” in AI development until more robust safeguards are put in place, expressing his concerns about the potential consequences of the ongoing rapid advancement in machine intelligence.

In the United States, Democratic Senator Bernie Sanders has proposed a temporary halt to advanced AI development and has introduced legislation to ban artificial superintelligence.

President Donald Trump has dismissed requests to halt AI development, expressing on Thursday his worry that the US might find itself “in a precarious position” if it does not succeed in the AI race.

Anthropic’s report marks the company’s inaugural threat intelligence report of the year, contributing to the growing concerns regarding the potential misuse of increasingly advanced AI systems.