This Hacker News article showcases an AI Liability Tribunal prototype that puts autonomous agents inside a fictional liability tribunal to make their communication observable. The demo puts agents with different roles, incentives, professions, and communication habits into a simulated legal setting, allowing jurors to choose whom to address and what to ask. The case is based on the fictional C-91 Incident, where an Aster Vale robotaxi struck and killed a cyclist.

The tribunal asks whether Aster Vale was criminally negligent in deploying the autonomous vehicle system. The default run is deterministic and offline, producing JSON results, Proto Link traces, a public transcript, and standalone interactive HTML reports. All agents and jurors are fictional, serving as a software experiment rather than a legal analysis.