Anthropic Blocks AI Use That Could Aid Bioweapons

Artificial intelligence company Anthropic said it has blocked multiple attempts to use its technology in ways that could support the development of biological weapons, according to a report it published.

The company behind the Claude assistant said it disrupted dozens of instances of potentially malicious use of its AI models from December 2025 onwards. Anthropic said some of the examples involved what it described as “biological misuse”.

The report sets out five case studies in which Anthropic’s models were used in ways the company said “could support biological weapons development”. It said it took action such as banning accounts linked to the cases, while adding that it could not establish the users’ intentions.

Blocking Suspected Misuse

Anthropic said it identified and disrupted activity it judged as high risk. The company did not provide details on how many accounts it banned in total, but said the cases involved different forms of engagement with its models.

Across the case studies, Anthropic characterised the potential harm as lying in how information generated through the system could be used to assist biological research linked to weapons development.

Case Studies Involving Biological Research

In one example, Anthropic said a researcher outside the United States accessed Claude from a region where the assistant is not supported. The company said the person used the model while working on highly pathogenic avian influenza, also known as bird flu.

Anthropic said the research focused in part on how the virus adapts to mammals. It said the topic was dual-use, explaining that knowledge about viral adaptation could help scientists detect naturally emerging variants with pandemic potential, but could also produce information that might be used to make a virus more dangerous.

In the other case studies, Anthropic described additional uses it said could support biological weapons development, based on the way the model’s outputs were being sought and applied.

Limits On Determining Intent

Anthropic said it banned accounts associated with the reported cases. The company stressed that it could not determine whether the researchers intended to cause harm.

“We do not assert that they intended harm,” the company said in its report.

The latest publication follows earlier efforts by Anthropic to outline risks. The company said it has previously issued three threat assessments describing malicious uses of AI.

Reaction After Coxon Warnings

The new report was released days after former Anthropic and OpenAI researcher Jacob Coxon spoke publicly about risks from more capable AI systems.

Coxon told CNN that, if capabilities expand, the systems could cause “extreme havoc”, including through hacking critical infrastructure and supporting the creation of “extinction-level bioweapons”.

His comments come as lawmakers in Washington have increasingly urged stronger government oversight and new safety rules for AI development.

Join the discussion?

11 September 2026