Researchers from the IC3 consortium have published a study titled Crypto x AI, AI x Crypto, challenging many popular claims about blockchain's benefits for artificial intelligence. The authors emphasize that meaningful integration of these technologies remains in early stages, and widespread narratives often outpace empirical evidence.

The work divides the topic into two directions. First – applying AI to crypto systems: here, artificial intelligence already shows practical value, helping analyze transactions, detect fraud, identify vulnerable smart contracts, and process on-chain events. Second – using crypto technologies to improve AI: researchers find the most tangible benefits in zero-knowledge proofs and trusted computing, while ideas like decentralized AI governance or distributed infrastructure remain largely experimental.

Special attention is given to common misconceptions. Blockchain cannot independently distinguish AI-generated content from human-created work – it merely timestamps digital artifacts without verifying their origin. Similarly, recording model outputs on-chain does not explain why a decision was made or guarantee training integrity. Researchers also dispute the claim that crypto wallets make AI agents truly autonomous: while wallets automate payments, they do not free operators from controlling the underlying infrastructure.

Regarding payments, the authors acknowledge blockchain's potential for censorship-resistant, neutral transactions but insist the industry must quantitatively demonstrate advantages over traditional financial systems. As for decentralized compute networks (DePIN), they may be cost-effective for small tasks but struggle with large workloads due to latency and bandwidth constraints compared to cloud providers. The study's core conclusion: the market must separate concept demos from proven utility – crypto technologies can enhance specific elements of AI infrastructure but do not automatically solve foundational challenges of trust, transparency, and computational efficiency.