Anthropic bans accounts after detecting attempts to use Claude for dual-use biological research
Anthropic reported that it blocked several researchers who tried to employ its Claude chatbot for biological work that could aid weapon development, resulting in account bans and tighter safety measures.
Anthropic said it had identified and disrupted multiple attempts this year by researchers to use its Claude AI model for biological investigations that could be repurposed for weapons. In a 154-page report titled “Detecting and Countering Misuse of AI,” the company described five case studies in which users sought assistance with projects such as gain-of-function studies on the chikungunya virus, bird-flu and orthopoxvirus research, and work on non-transmissible venoms and toxins. The company blocked the requests, banned the associated accounts and reported the findings to government authorities and other AI firms.
"We banned all associated accounts, worked with partners to take down the relay networks that evaded regional blocks and shared our findings with affected AI labs and government authorities," Anthropic said.
The report noted that while the same scientific information can support vaccine and treatment development, it also carries a risk of misuse. Anthropic highlighted that some actors attempted to bypass regional restrictions by routing rejected prompts through third-party platforms.
The company also cited an Iran-linked threat actor that used Claude to gather open-source data for targeting recommendations against U.S. naval forces, leading to another account ban and the sharing of threat intelligence with authorities.
Separately, a former Anthropic researcher publicly warned of a high probability of catastrophic outcomes from AI, but Anthropic did not comment on those statements.