Can AI and blockchain trust each other?
What happens when AI can act on a blockchain without waiting for a human to approve every move? As AI agents become more autonomous, they could manage blockchain accounts, move assets, and even participate in network governance, raising a fundamental question: how much should they be trusted?
Image by Magnific
A review published in Blockchain examined this two-way trust problem, looking at both how blockchain systems can control AI agents and how AI can safely participate in blockchain networks.
The review examined 127 academic papers, 20 industry projects, and 70 Ethereum proposals and standards. It found that the challenge goes beyond simply giving AI access to blockchain systems. Developers must establish clear limits on what agents can do, verify their actions, and ensure they follow the user's original intent.
The problem has a second side. AI could also become involved in security auditing, network consensus, and governance, potentially giving agents a role in how blockchain systems operate and evolve. Yet research in these areas remains fragmented, with no unified security framework for treating AI as a participant at the protocol level.
The standards supporting AI agents are also still developing. As of July 18, 2026, only three of 13 standards directly targeting AI or agents had reached final status, while the other ten remained drafts.
Another unresolved question is whether an AI agent can reliably turn a user’s goal into the correct on-chain action. The review found limited formal analysis of intent-based execution, leaving uncertainty over how systems can ensure that what an agent does is exactly what its user intended.
Several technologies could help establish trust, including cryptographic proofs, challenge-based systems, and trusted hardware. However, they come with different assumptions, costs, and levels of readiness, and none provides a complete answer.
The researchers identified nine areas requiring further work, including verifying AI decisions, enforcing delegated permissions, establishing reliable agent identities, and defining how humans and AI should share roles in blockchain governance.
As AI agents become more autonomous, the challenge may be less about what they can do than whether blockchain systems are ready to trust them.
DOI: 10.55092/blockchain20260008