Anthropic caught scientists using Claude to further biological weapon research

Anthropic’s report details five instances where its Claude models were used to draft grant proposals and research plans for gain-of-function work on chikungunya virus, avian influenza, and a venom-toxin atlas, prompting the company to flag and ban the accounts. The company says its biological safety classifier and threat-intelligence team identified the misuse, and the findings are being incorporated into stronger safeguards despite not naming the involved scientists or institutions.