Anthropic says Claude aided bird flu research; experts question added capability
Anthropic reported a researcher used Claude to plan avian flu work and select mutations. Experts say the report lacks detail on whether the AI added new capability.
The story in focus
Anthropic reported that a researcher used Claude to plan studies, analyze data and choose experiments on highly pathogenic avian influenza, including mutations that could help the virus adapt to mammals. The company banned the account in May after the researcher evaded regional access restrictions, and experts told Science News the report lacks detail on what Claude actually added.
- Anthropic described the case in a threat intelligence report released September 10.
- The researcher used workarounds to reach Claude from a region where it was not supported, and Anthropic banned the account in May.
- Anthropic said it could not determine whether harm was intended because legitimate and dangerous biological research can look similar.
- Experts say the report does not show whether Claude merely sped up existing work or added a new scientific capability.
- Anthropic reported that the bird flu research was one of five cases in its report involving uses that could support biological weapons development.
What Anthropic reported
Anthropic's threat intelligence report, released September 10, says a researcher relied on Claude to plan studies, analyze data and choose experiments involving highly pathogenic avian influenza. The work sought to identify and introduce mutations that could help bird flu viruses adapt to mammals and transmit more easily between them.
Anthropic says that, to its knowledge, no private company had previously publicly shared evidence of potential misuse of its platform for biological weapons development. The company did not respond to Science News questions about why it considered the case significant, how it assessed Claude's role, or whether any proposed experiments were performed.
Access and the May ban
According to Anthropic, the researcher accessed Claude from a region where the service was not supported and used workarounds to get in. The material shared with Claude suggested the group probably had physical access to relevant virus samples, the company said.
Anthropic banned the account in May for evading access restrictions. The company said legitimate and dangerous biological research can look alike, so it could not tell whether harm was intended.
What experts question
Influenza researcher Thomas Friedrich of the University of Wisconsin–Madison says studying how avian influenza viruses might pose a pandemic risk is a normal part of virology. He says the company's own account suggests Claude did not meaningfully add scientific capability.
Biodefense researcher Gregory Koblentz of George Mason University told Science News there is no evidence the research was connected to biological weapons development. Kevin Esvelt of the MIT Media Lab said none of the five cases in the report represented a competent attempt at weaponization.
Friedrich notes a large gap between identifying mutations that might adapt a virus to humans and a harmful biological outcome. Koblentz says turning a design into a threat still requires modifying or synthesizing the organism in a lab, along with facilities, containment, funding and trained personnel.
Did Anthropic show that Claude enabled new biological capability?
No. Experts told Science News the report lacks detail on whether Claude sped up work the researcher could already do or added a new capability. The company's account suggests the model did not meaningfully add scientific capability, Friedrich said.
Sources & further reading
Help keep the record accurate. Send a correction.
