Anthropic details global Claude misuse in new threat report

Anthropic has published a new threat report summarizing misuse of Claude that it says it disrupted over the past eight months. The report covers attempts tied to bioweapon concerns, espionage, rocket guidance, malware evasion, and other harmful uses of the model. The company said it identified Chinese labs involved in distillation efforts and described cases where Claude was allegedly served to customers in place of a proprietary model. It also noted several biologyrelated cases that raised safety flags, while saying it did not claim the users intended harm. Other examples in the report included a rocket guidance workflow in Yemen and surveillance software work linked to Mali’s spy agency. Anthropic said the cases involved models at Opus level and below, and that more capable systems could create even greater safety challenges.