AI “judges” locked into on‑chain model, prompt, and data‑source commitments could restore trust and scale prediction markets by providing neutral, manipulation‑resistant resolution for ambiguous real‑world events while keeping human governance only at the meta‑level.