Anthropic blocks possible attempt to use AI to make biological weapons

London, Sep 11 — AI firm Anthropic has uncovered and stopped attempts to exploit its AI technology for "malicious activity", which could potentially aid the creation of biological weapons. The discovery is detailed in the company's latest threat intelligence report, which highlights the growing concerns surrounding AI's potential risks to humanity.

The report reveals that malicious use of Anthropic's Claude Haiku, Sonnet, and Opus models was stopped between December 2025 and August 2026. The incidents include cases involving cyber espionage, surveillance, scams, fraud, and the development of biological weapons. Despite the serious risks posed by misuse, Anthropic emphasizes the potential benefits of AI, such as the development of vaccines and cures for diseases.

CEO and head of threat intelligence Jacob Klein stresses the nuanced nature of the situation, noting that the misuse often involves scientists attempting to create tools for conventional weapons, rather than outright weaponized biological agents. The report also mentions that Chinese AI companies have been trying to replicate Claude's capabilities.

Anthropic has taken steps to address these concerns by incorporating its findings into its processes and sharing intelligence with authorities and industry partners. The company's warning comes as top AI researchers and lawmakers call for caution and potential regulation in the rapidly advancing field of AI.