Buterin argues AI should support DAO voters, protect privacy, and solve attention limits without centralizing control.
Ethereum co-founder Vitalik Buterin has argued that decentralized governance faces a basic human limit. In a recent post on X, he warned against handing power directly to artificial intelligence systems. Instead, he proposed using personal AI agents to support democratic decision-making.
Buterin Rejects AI Rule, Proposes Personal Agents for DAOs
According to Buterin, handing power directly to AI systems would create new risks. Weak systems could flood governance with poor judgments. On the other hand, more advanced systems could centralize influence in ways that undermine decentralization itself.
For that reason, the co-founder rejects the idea of “AI as government.” Instead, he sees AI as support infrastructure for human voters. He stated that decentralized organizations require thousands of decisions across legal, technical, and financial domains. As such, most participants lack the time or expertise to assess each proposal.
Delegation has been a common practice, especially in DAOs built on Ethereum. However, when many people delegate their tokens to the same few representatives, those representatives end up with most of the decision-making power. After delegating, regular token holders usually stop having direct control over votes.
Buterin suggests that instead of people voting on every small issue themselves, they could use a personal AI assistant to do so. That assistant would learn their views from what they have written or said before.
However, if uncertainty occurs, the agent would request direct human input. In that setup, final authority remains with the individual, not the software.
He also believes decision-making should involve better public discussion. AI systems could convert private inputs into standardized public formats without exposing personal data. They could also group similar viewpoints, allowing participants to respond to refined summaries.
Ethereum Co-Founder Pushes Black-Box AI Model for Sensitive Decisions
Buterin also discussed “suggestion markets.” In such systems, anyone could submit proposals or arguments. AI systems would place bets on tokens linked to each submission. If governance mechanisms accept a proposal or treat it as valuable input, token holders receive rewards.
Financial signals would help surface stronger ideas without relying on centralized review committees.
In Buterin’s view, decentralized systems can handle sensitive decisions without exposing private information. He pointed to multi-party computation as a possible solution. This method allows several people’s inputs to be processed together while keeping the actual data hidden from everyone involved.
He explained that a person could place their personal AI assistant inside a secure environment, similar to a black box. Inside that system, the AI could review confidential information and form a judgment.
However, it would only release the final decision. Neither the user nor other participants would see the raw private data. At the same time, no one would have access to the personal AI system’s internal contents.
Buterin stressed that privacy must sit at the center of such governance models. As participants rely more on personal data and larger inputs, risks increase if protections are weak. For that reason, anonymity becomes essential. Tools such as zero-knowledge proofs can allow someone to prove eligibility or validate a vote without revealing identity.
Source: https://www.livebitcoinnews.com/vitalik-buterin-says-daos-struggle-with-human-attention-scarcity/
