A research survey published Monday by the Initiative for CryptoCurrencies and Contracts (IC3) challenges many claims about blockchain's ability to solve artificial intelligence's toughest problems. The paper, authored by researchers from Cornell, Carnegie Mellon, Princeton, Yale and ETH Zurich, argues that crypto's role in granting AI agents autonomy, detecting AI-generated content, and mitigating algorithmic bias is frequently overstated.
One key misconception addressed is the belief that giving AI agents a crypto wallet makes them fully autonomous. The authors note that such agents can automate transactions without human approval loops, but they remain dependent on human operators and traditional infrastructure. "AI systems do not become more intelligent by possessing a wallet. Nor do they become more resistant to human manipulation or shutdown," the paper states.
Coinciding with the study, Consensys announced early access to MetaMask Agent Wallet, a non-custodial wallet designed for AI agents. Founder Joe Lubin said machine intelligences will increasingly transact on crypto rails. Separately, Robinhood confirmed plans to let users delegate cryptocurrency trading to AI agents, starting with a beta in equities.
IC3 also examined the idea that blockchains can distinguish human- from AI-generated content. While blockchains can timestamp and register digital artifacts, they cannot verify how content was originally created. An external classifier would be needed, and any mistake it makes would be permanently preserved on-chain, rendering the record-keeping function of limited utility for this problem.
On algorithmic bias, the researchers found decentralization does not address the root cause. Bias arises during model training, and decentralized governance may increase transparency but not correct the bias itself. The paper was edited by Carnegie Mellon's Giulia Fanti and Cornell Tech's Ari Juels, who is also chief scientist at Chainlink Labs.
