Biosecurity experts warn AI could lower barriers to harmful biology

In short: Biosecurity experts are raising alarms that AI tools are starting to make biological design easier, even though there is no evidence an AI-made human pathogen is imminent.

What's going on

A recent demonstration added fuel to the debate. Researchers at Stanford University and the Arc Institute said they designed and manufactured 16 working bacteriophages, which are viruses that infect bacteria, not people. The phage genomes were not found in nature, showing AI can help design living systems, but it did not show AI can create a dangerous virus that infects humans.

Another concern comes from reported misuse of AI chatbots. Anthropic said it found five cases, between December 2025 and August 2026, where people used Claude in ways that could support biological weapons development. The company said it suspended access in those cases, but it did not say a weapon was built.

Experts say the bigger issue is the direction of travel. AI could help people with less training work through hard biology problems, like a calculator that makes math faster. At the same time, DNA synthesis (ordering custom DNA like placing a mail order) and lab automation could make it easier to turn digital designs into real lab work. That can raise risks from deliberate misuse and from accidents in labs that already vary widely in oversight.

What to watch

Researchers and policymakers are calling for stronger testing and safeguards for AI models, independent assessments, and better screening of DNA synthesis orders. Some also want closer monitoring of automated labs and clearer rules that tie warning signs to specific actions. The White House announced a voluntary accord in October encouraging AI companies to improve internal monitoring and oversight, but much of the follow-through is left to the companies.

Source: NYTimes