Failing to act on warning signs: from Letby to AI

AI Summary

Concerns are rising about artificial intelligence posing existential threats, including the potential creation of biological weapons. Researchers and AI company insiders have highlighted safety issues and the need for surveillance systems to mitigate biological threats possibly exacerbated by AI misuse.

Might artificial intelligence indeed kill us all? The chances are greater than 10%, according to one expert. Another researcher has publicly quit the AI company Anthropic over safety fears.1 This news arrives around the same time that Anthropic warns us that people are attempting to use AI to create biological weapons (doi:10.1136/bmj-2026-100871).2 The threat from such weapons is real and is a major concern for ministries of defence around the world. A BMJ analysis paper earlier this year outlined how a surveillance system for biological threats might be constructed (doi:10.1136/bmj-2025-086457).3The warnings are being served. Is it wise to ignore them? One view is that safety concerns about AI are being exaggerated to allow the leading AI companies to maintain their competitive advantage and maintain their influence over regulation. That may be true to some degree, but is it sensible to park an existential threat when the evidence suggests that our...