BMJ 2026; 394 doi: https://doi.org/10.1136/bmj-2026-100871 (Published 16 September 2026) Cite this as: BMJ 2026;394:e100871
- Canqi Li
- The BMJ
Scientists have used, or attempted to use, the AI tool Claude five times for biological research that could support the development of bioweapons, the company that owns it has claimed.
Details of the five cases were revealed in a safety report by the AI company Anthropic.1
Examples it cited included seeking Claude’s assistance in research projects to genetically engineer more infectious strains of the chikungunya virus and a mammalian adaptation of highly pathogenic avian influenza.
Other attempts were to assist in a grant application to explore immune evasion or virulence related genes in orthopoxviruses and plans to redesign deadly toxins.
Anthropic said that this was the first public disclosure by a private company of evidence that its platform may have been misused in ways relevant to developing biological weapons. “Biological misuse is one of the most serious risks of frontier AI models,” …
https://www.bmj.com/content/394/bmj-2026-100871#main-content
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.