Anthropic has released a threat intelligence report detailing how its Claude AI models were misused for activities connected to cyberattacks, surveillance, conventional weapons software, and research that could support biological weapons development.
The report covers cases detected and disrupted between December 2025 and August 2026. Anthropic said it identified and blocked multiple accounts involved in these activities.
In the biological area, the company described five case studies in which users, including scientists, tried to use Claude in ways that could support bioweapons-related work. Some involved attempts to hide the purpose of the research or bypass geographic restrictions. One example involved help with a grant application related to gain-of-function research on a virus. Anthropic stressed that biological misuse remains one of the most serious risks of frontier AI models and said it has strengthened safeguards as a result.
The report also detailed cyber operations. Actors linked to a Russia-based group used Claude in phishing, malware evasion, and related activities targeting institutions in Ukraine and elsewhere. Other cases involved scams, fraud, and influence operations.
China-linked activity appeared in several areas. These included surveillance efforts and attempts by some China-based companies or labs to extract or distill capabilities from Claude models for their own systems. Anthropic also noted conventional weapons-related work, such as software development for drones, missiles, and related systems, involving actors associated with Russia, China, and Yemen.
Most of the misuse involved Claude Haiku, Sonnet, and Opus models. Anthropic said it banned the accounts, disrupted the operations, and improved its detection and safety systems.
The company stated that these examples represent the most notable threat activity identified so far, not typical everyday misuse. It published the findings to increase transparency about real-world risks as AI models become more capable.
Reference links
- https://www.bbc.co.uk/news/articles/cx2zrrpkx20o
- https://apnews.com/article/anthropic-ai-threat-bioweapon-russia-00266dca90e4f8853f669648998d3bda
- https://www.theguardian.com/technology/2026/sep/10/anthropic-report-details-ai-misuse
- https://www.cnn.com/2026/09/10/health/anthropic-bioweapons-report
