Anthropic has blocked scientists who used its Claude artificial intelligence models in ways that could have supported the development of biological weapons, the company has said.
The cases are among five detailed in a new threat intelligence report covering activity disrupted between December 2025 and August 2026. Anthropic said the examples demonstrated potential capability, but did not prove that any of the researchers intended to create biological weapons or that such weapons were developed. ([anthropic.com](https://www.anthropic.com/threat-intelligence-report-september-2026))
The company has not identified the scientists, describing them as working researchers whose motives it could not establish. Their accounts were banned after investigations, with the findings used to strengthen Claude’s safeguards and systems for detecting suspected misuse.
Anthropic said biological misuse was one of the most serious risks posed by advanced AI models. One case involved virologists working on chikungunya gain-of-function research under a state-sponsored grant, while another concerned a researcher planning experiments involving the adaptation of avian influenza to mammals. ([anthropic.com](https://www.anthropic.com/threat-intelligence-report-september-2026))
The report said newer models, including Claude Fable 5, had been launched with tighter controls limiting access to a broad range of dual-use biological research queries. Such research can have legitimate medical applications, however, and Anthropic said the same knowledge could potentially be used to develop a vaccine or treatment.
“Although evaluations are useful in that they provide evidence of capability, they cannot concretely demonstrate that such capability would ever be used to develop biological weapons in the real world,” the company said.
Claude accounts linked to surveillance and propaganda
Anthropic’s report also described the removal of accounts involved in influence operations and surveillance. Three Iranian state-aligned operations allegedly used Claude to produce propaganda, imitate independent news sources and distribute material across platforms including X, Instagram and TikTok.
In one case, the company said an Iran-linked actor used Claude to produce targeting recommendations concerning US naval forces in the region and to design software for a domestic mass-surveillance system.
Anthropic also identified surveillance activity involving actors in China and West Africa. Three accounts aligned with a Chinese municipal security service allegedly used Claude for monitoring and transnational repression, including the profiling of overseas activists, dissidents and civil society organisations. ([anthropic.com](https://www.anthropic.com/threat-intelligence-report-september-2026))
The company said Claude had in some instances been used as a substitute for an engineering or analytical workforce, helping operators build surveillance systems, compile dossiers and generate regular intelligence reports.
Anthropic said it had disrupted each case described in the report, banned the accounts involved and shared information with authorities and industry partners where appropriate. The report said the activity demonstrated how advanced AI systems could be used for harmful purposes by criminals, suspected state-backed groups, spyware vendors and propaganda institutions.
