Who Pays When the Machine Is Wrong?

This talk maps who is legally liable when autonomous code causes loss: an AI agent signs a bad transaction, an oracle misreports, an upgrade freezes funds.

The main finding is that liability follows control. Admin keys, upgradeability and frontends are also the points where the law attaches responsibility, so design choices are liability choices. Agents make this sharper: when no human authored a decision, agency and negligence rules break down.

Our goal is to raise builders' awareness of these risks early, while they can still shape them through design. Attendees leave with a framework to locate their own exposure and concrete steps to reduce it before launch. This is timely because agents are going onchain now, while the rules deciding who pays are being set right now.

Speakers: Patrick Storchenegger, Raphael Baumann

Track: Users, Builders, and Agents