AI company Anthropic says it has identified five cases in which users employed its Claude artificial intelligence models for activities that could potentially support biological weapons research.
The company disclosed the cases in its latest threat intelligence report, published on Thursday, covering malicious or potentially harmful activities detected and disrupted on its platform between December 2025 and August 2026.
According to Anthropic, the five cases involved research areas including chikungunya virus, highly pathogenic avian influenza, orthopoxviruses, venom peptides and toxins.
The company said it banned the accounts involved and used information from its investigations to strengthen its safety measures, enforcement systems and threat intelligence capabilities.
Anthropic, however, stressed that it was not claiming that the researchers intended to cause harm. It said it would not disclose the names of the institutions, countries or specific biological agents and techniques involved because doing so could potentially expose the researchers or laboratories to harm.
The company said the cases demonstrate the difficulty of separating legitimate scientific research from activities that could potentially be misused.
Anthropic described biological misuse as one of the most serious risks associated with advanced AI models, noting that information useful for developing harmful biological agents could also contribute to the development of vaccines and treatments.
Chikungunya Research Request Blocked
In the first case, Anthropic said its biological safety system blocked a request related to preparing a grant application involving gain-of-function research on chikungunya virus.
A subsequent investigation allegedly found that a reseller platform was providing Claude access to researchers in regions where Anthropic does not officially offer its services.
The company said the platform also redirected requests rejected by Claude to other AI models with less restrictive safeguards.
Anthropic said it subsequently banned the accounts involved, worked with partners to shut down the relay networks and shared information from the investigation with other AI companies and government authorities.
Research Into Highly Pathogenic Bird Flu
In a second case, Anthropic said a researcher outside the United States used Claude over several weeks while planning research involving highly pathogenic avian influenza and its adaptation to mammals.
The researcher reportedly used the AI system for activities including research planning, data analysis, interpretation, experiment prioritisation and editorial assistance.
Anthropic said its safety measures restricted the researcher to less capable Claude models, limiting the AI’s role mainly to clerical work, research ideas and study design.
However, the company said the case provided evidence of active research programmes involving pathogens with the potential for enhanced pandemic risks.
Orthopoxvirus Research
The third case involved a grant application for orthopoxvirus research at a state-associated infectious disease laboratory.
Anthropic said a reseller relay serving more than a dozen customers had used Claude to assist in drafting the application.
The company said the research focused on how orthopoxviruses interact with the immune system and that the application was produced using Claude Opus 5 in about an hour.
Venom and Toxin Research
The fourth and fifth cases involved research into venoms and toxins.
Anthropic said one researcher used Claude to develop an atlas of venom toxin peptides and a generative system intended to optimise toxin characteristics, with the stated objective of developing therapeutic molecules.
In another case, a researcher used Claude for projects involving the computational redesign of toxins under a national public research programme.
Anthropic said that researcher deliberately concealed the identities of some toxins and viral proteins in progress reports.
The company said both accounts were banned in May 2026 for violating its Supported Regions Policy.
Anthropic Warns of Dual-Use Risks
Anthropic said the cases demonstrate the regulatory challenges surrounding AI-assisted biological research because many scientific applications are dual-use, meaning the same knowledge can have both beneficial and harmful applications.
The company cautioned, however, that the findings should not be interpreted as evidence that Claude is currently enabling an imminent biological threat.
Instead, Anthropic said the cases showed that significant dual-use research activities were associated with state actors considered to be of concern.
The report forms part of Anthropic’s wider disclosure on the misuse of its AI systems.
The company said its investigations during the eight-month period also uncovered cyber operations, surveillance, influence operations, scams and fraud, conventional weapons development and attempts to illicitly distil its AI models.
The report follows concerns previously raised by former Anthropic researcher Jacob Coxon, who resigned from the company and warned publicly about the potential risks posed by increasingly capable AI systems.
Anthropic’s latest report also disclosed a Russia-linked cyber espionage operation in which Claude was allegedly used to automate parts of attacks targeting organisations in Ukraine and Europe.
Source: PUNCH
