Anthropic: Use of AI that could help develop biological weapons blocked

Anthropic, the company behind the Claude AI model, announced it blocked malicious attempts to use its models, including research aimed at increasing the infectivity of the chikungunya virus. The company published a report documenting misuse between December 2025 and last month, including cyberattacks and influence operations. Meanwhile, a senior researcher resigned, warning of an uncontrolled development race.

Anthropic, the company operating the Claude AI model, announced yesterday (Thursday) that it blocked malicious attempts to use its models, including research that could have helped develop biological weapons. The announcement followed a report published by the company, documenting misuse between December 2025 and last month. The report includes documentation of cyberattacks, surveillance, and influence operations on social networks. One major case involved a request to Claude to help write a research funding application aimed at increasing the infectivity of the chikungunya virus and its ability to evade the immune system. The company noted that such research could be used to develop vaccines and treatments, but could also enable the development of a more dangerous virus. The company emphasized that the cases detailed are the most prominent and innovative threats identified so far, and that protective measures have been tightened in new models. Meanwhile, a senior researcher at Anthropic, Jacob Coxson, resigned and announced he would leave the AI field entirely. Coxson stated he is concerned that tech giants are rushing to develop uncontrollable systems with self-improvement capabilities, which could spiral out of control and destroy humanity. In June, the company announced it had blocked access to its most advanced models following a US directive.