Leon Staufer

Research

I work on AI governance and safety, with a focus on evaluations and AI agents. Right now, I’m researching emergency shutdown mechanisms for advanced agents at SASH. Previously, I’ve worked on documenting deployed agents and making AI audits and evaluations easier to interpret.

I’m interested in how we can understand and govern increasingly autonomous AI systems. What do frontier evaluations tell us, what developers should disclose, and how people can retain meaningful control? Fundamentally, I care about how AI changes who has power and who gets a say. I’m open to collaborations, so please get in touch if you’d like to work together.

You can see my publications on Google Scholar as well.

Publications

* Equal contribution. These authors share first authorship.

Invited talks

Selected media