2026-11-05 –, Jasmine Stage 4
Agent-to-agent commerce is coming: agents negotiating, paying, and contracting with each other at scale. That world needs governance: ways to curate and certify which agents can be trusted, and agent juries to resolve disputes when they arise. Based on real experiments, this talk maps out that infrastructure and the hard problems in building it.
An economy where AI agents transact directly with each other (negotiating, paying, contracting) is arriving fast, but the governance infrastructure isn't there yet. Many disputes will happen at machine speed, for amounts too small to justify a human court case. This talk maps out what that infrastructure needs: a way to certify which agents can be trusted to transact, and a way to resolve disputes when parties inevitably disagree: agent juries ruling on agent-to-agent cases.
But it's not that simple. In real experiments, different frontier models ruled differently on the exact same disputes: Claude was nearly 5x more likely to rule in the consumer's favor than ChatGPT. Same facts, same rules, different model, different verdict. Plus early signs that agents can carry human-like biases into their decisions. What does neutral dispute resolution look like for agents, and what mechanisms get us there?
These are critical challenges anyone building this infrastructure will have to solve.
Federico Ast is the founder and CEO of Kleros, a decentralized dispute resolution protocol built on Ethereum. He holds degrees in Economics and Philosophy and a doctorate in Business Management, and teaches blockchain and the future of law to over 120,000 students on Coursera. His work sits at the intersection of blockchain, game theory, institutional design, legaltech, and AI.