Artificial intelligence company Anthropic says it has identified and shut down accounts that used its Claude models in ways that could support the development of biological weapons, in a threat intelligence report published on Thursday.

The report sets out five case studies in which users “circumvented controls” blocking access from certain regions and “engaged in other efforts to obfuscate the purpose of their research to evade our safeguards,” CNN reported.

The examples include possible gain-of-function work — one user sought help drafting a grant application for research on the transmissibility and immune evasion of the chikungunya virus — along with a researcher outside the United States working on how bird flu adapts to mammals, research into orthopoxviruses (the family that includes smallpox and mpox), and the computational redesign of venom toxins.

Reviewing 30 days of activity, the company said it found about 35 “distinct research efforts” of potential concern. It acknowledged it could not establish whether those involved “intended harm” or were doing legitimate science, and said the individuals appeared to be “working scientists.” It did not name the institutions or countries involved.

Beyond biology

The report covers misuse of the Claude Haiku, Sonnet and Opus models between December 2025 and August 2026. It says none of the cases involved Claude Fable or the company’s Mythos-class models, apart from a single instance of distillation — training a smaller model on the output of a larger one.

Claude was also used by actors linked to a Russia-based cyber espionage campaign and by an Iranian propaganda institution, Hiru News reported. Other cases ranged from fake dating apps and hotel Wi-Fi scams to surveillance tools built to identify dissidents. Anthropic separately accused Chinese AI firms of trying to replicate Claude’s capabilities.

“Sophisticated attacks no longer require sophisticated attackers,” the report says, arguing AI has “collapsed the labor and tooling gap” between state-backed operations and lone operators.

Jacob Klein, who heads threat intelligence at the company, told the New York Times the judgement calls were difficult. “You are not seeing someone in a comic book kind of way say, ‘Hey, I want to build a biological weapon to kill everybody,’” he said.

Anthropic called biological misuse one of the “most serious risks” of frontier AI models, noting that “the same information that can be used to develop a biological weapon could also be used to develop, for example, a vaccine or a cure for a disease.”

Industry disclosures and political reaction

Such reports have become routine as AI firms seek to show they can detect misuse. Google said on Tuesday that someone had tried to use Gemini to obtain a “complete, step-by-step technical guide for synthesizing weaponised biological agents.”

The disclosures land amid sharpening political argument. US Senator Bernie Sanders has called for a pause on advanced AI development and a ban on artificial superintelligence — a position he set out days after an Anthropic researcher resigned with a public warning that AI firms are “gambling with our lives.” President Donald Trump has rejected that framing, saying on Thursday he was concerned that “if we don’t win AI, we’re going to be put in a very bad position.”

Not reported

Neither report says which countries the flagged researchers were working in, whether any were referred to national authorities, or how Anthropic distinguishes dual-use academic work from weapons development in practice. Neither gives the number of accounts banned.