Anthropic says it has blocked accounts using its Claude artificial intelligence models for research that could support the development of biological weapons.

The company’s threat intelligence report, published on September 10, identifies five biological research cases among AI uses disrupted between December 2025 and August 2026. The findings form part of a broader investigation into potentially harmful applications of its technology.

Anthropic said the users involved were working scientists and stressed that it was not alleging they intended to cause harm. It withheld their names, institutions, and countries to protect those concerned.

The concern centres on dual-use research: scientific knowledge that can support legitimate medical work but could also be applied to weapons development. The company said some users had obscured their research purposes or bypassed regional access restrictions.

Beyond biology, the report described Claude’s use in cyberespionage, propaganda, surveillance, scams, and software for conventional weapons. Cases included fake dating services, hotel Wi-Fi scams, and systems intended to identify dissidents.

Anthropic said it had banned accounts, strengthened safeguards, and shared intelligence with authorities and industry partners where appropriate. It described the published cases as notable examples rather than a representative picture of all Claude use.

The findings do not establish that biological weapons were produced or that the scientists involved intended to develop them.