Anthropic says its models were misused for biological weapons research, surveillance and cyber attacks

On September 10, 2026, Anthropic, an artificial intelligence company, released a detailed report revealing instances of its AI models being misused for potentially harmful purposes. These included developing biological weapons, conducting cyber attacks, influence operations, and surveillance.

The report came in response to a growing concern about the potential risks posed by increasingly powerful AI models. It detailed how individuals had attempted to exploit Anthropic's Claude AI for malicious activities, and how the company was able to detect and stop these operations.

One example provided in the report was a prompt entered into Claude that sought assistance in applying for a grant to research the chikungunya virus, a mosquito-borne disease known for causing severe, debilitating symptoms with no licensed treatments available. The report noted the concern raised by the institutional affiliation associated with this grant application, though it did not provide further details.

Anthropic emphasized that it had disrupted every operation mentioned in the report. They have implemented several internal checks to manage potentially problematic prompts and are continually refining their methods to prevent misuse of their services.

The company also acknowledged the challenge in balancing the potential for beneficial scientific research with the need to protect against misuse. They expressed hope that AI would contribute to advancements in biological science and lead to new medical treatments.

The report has sparked mixed reactions. While some view it as a necessary step to ensure safety, others argue that it may generate unnecessary panic. The debate has also drawn attention to the growing political scrutiny around AI, with some lawmakers proposing legislation to limit AI research.