Anthropic details Claude misuse in new threat report — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic published its September 2026 threat intelligence report on Sept. 10, detailing disruptions of Claude misuse across seven harm areas: cyberattacks, surveillance, influence operations, conventional weapons, biological misuse, scams and fraud, and illicit distillation.
- Anthropic's report covers misuse detected between December 2025 and August 2026, and the company states it disrupted every operation mentioned in the document.
- The New York Times reports Anthropic shut down several research efforts by scientists using Claude because the company could not determine whether the work was legitimate or aimed at developing biological weapons.
- Politico reports bad actors in China and Russia are weaponizing Anthropic's AI, while Axios reports governments are turning to Claude to automate spying.
- NBC News reports state-linked scientists used Claude for risky virus research, and the Washington Examiner highlights Anthropic's alarm over AI models being misused for biological weapons development.
- Susan Zhang, a former Anthropic employee, publicly questioned the report's motives on X, writing: 'oh no chynaaa oh no biorisk... please nationalize us! or buy our IPO!'
Why it matters: Anthropic documented state-linked misuse from China and Russia — including biological weapons research and government surveillance — giving the company a concrete safety record to wield against competitors as AI labs face escalating biosecurity scrutiny. The company's choice to shut down work it could not verify as legitimate shows it is blocking suspicious users, not merely flagging them, setting a more aggressive posture than 'monitor and warn.'
Ask SkimNews
