Constructing a Decentralized AI Data Marketplace Enabled by a Blockchain-Based Incentive Mechanism

Abstract

As data increasingly becomes a key factor of production for artificial intelligence (AI), this paper proposes a blockchain-enabled, decentralized AI data-market framework. To address the long-standing problems of low transparency, high privacy risk, and misaligned incentives in traditional data trading, we design a layered hybrid consensus that combines Proof of Stake (PoS) with Practical Byzantine Fault Tolerance (PBFT), balancing economic security with sub-second finality. A token-based incentive model that weights data quality, volume, and staking risk is introduced to couple value discovery with the suppression of low-quality data. By combining symmetric encryption with proxy re-encryption, the framework allows data to be “usable yet invisible” while exposing a compliance interface for regulated auditability. A prototype deployed on 18 nodes achieves 4,750 tx · s⁻¹ throughput and 148 ms latency, with energy consumption far below Proof-of-Work (PoW) schemes—demonstrating performance, privacy, and ESG friendliness. This work provides a reproducible technical path and theoretical foundation for sustainable innovation in data-factor circulation and AI applications.