🦠 Biosecurity Futures | AI Meets Biology

​AI Meets Biology
What Frontier Models Actually Change

​Monday, 12 October 2026 · 18:15–19:45
ETH Zürich, room to be confirmed

​Every major AI lab now publishes commitments about biological misuse. Governments cite it in national security strategy. The claim is that these systems could help someone build a weapon they otherwise couldn't.

​The interesting question is whether that's true, and how anyone would know.

​Information has never been the only barrier to building a biological weapon. There are also materials, equipment, and skills that are hard to acquire and harder to write down. Which of those barriers a language model actually lowers is an empirical question, and a surprisingly difficult one to answer.

​This session looks at what "uplift" means precisely enough to argue about, how labs and evaluators try to measure it, and why safeguarding these systems turns out to be harder than it sounds. Expect open questions rather than settled answers.

​No biology background needed, and no machine learning background either.

​Please note this session is on a Monday, the only one in the series that isn't on a Wednesday.

​Biosecurity Futures is an introductory series on global catastrophic biological risks, hosted by the Zurich Biosecurity Hub at ETH Zürich.

​Biosecurity Futures Series Schedule:

1. ​Worse Than Nature: Why the Deadliest Pandemics Won't Be Natural Ones — 30 Sep

2. ​AI Meets Biology: What Frontier Models Actually Change — 12 Oct (Monday)

3. ​Mirror Life: A Risk We Could Still Prevent — 28 Oct

4. ​The Air We Share: Clean Air, Far-UVC and Defences That Work Against Anything — 11 Nov

5. ​Layers of Defence: Screening DNA, Spotting Outbreaks, Stacking Imperfect Safeguards — 2 Dec

​Open to students, researchers and staff from all departments and all universities, and to interested people from outside academia. Free.

​Subscribe to the calendar to get all five sessions.

​More at zurichbiosecurity.ch. Questions: info@zurichbiosecurity.ch