Coverage detailed biosecurity cases and the admission that newer models carry greater biological research risk
Anthropic published a threat intelligence report on September 10 covering attempted misuse of Claude for cyberattacks, influence operations, and weapons development. The report details five blocked biological research cases, including one tied to a military research institute working on chikungunya; a reseller serving those researchers regained access within days of a ban. Anthropic states it can no longer give its prior assurance that Claude models are too limited to meaningfully aid dangerous biological research, and has broadened the biological requests it restricts.