Vitalik Proposes Private AI Stewards to Scale DAO Governance and Curb Whales

Vitalik Buterin has proposed a radical technical fix for a core problem dogging many DAOs: too many decisions, too little meaningful participation, and growing power concentration among large token holders. In a post on X, the Ethereum cofounder outlined a roadmap that leans on personal AI agents — or “stewards” — to vote privately on users’ behalf and scale decentralized governance. Why AI stewards? Buterin argues people simply can’t follow or adjudicate the thousands of proposals that modern DAOs face. “There are many thousands of decisions to make, involving many domains of expertise, and most people don't have the time or skill to be experts in even one, let alone all of them,” he wrote. His solution: personal LLMs trained on an individual’s historical messages and stated values, which would act as delegated voters that reflect the owner’s preferences. Key elements of the proposal - Privacy-first voting: To prevent sensitive data from being exposed on-chain, these AI agents would run in secure environments such as multi-party computation (MPC) or trusted execution environments (TEEs). That lets agents evaluate private inputs (think job applications or legal disputes) without leaking them publicly. - Anonymity and anti-coercion: Buterin recommends zero-knowledge proofs (ZKPs) so users can prove voting eligibility without revealing wallet addresses or vote contents. That helps reduce coercion, bribery and “whale watching” where small holders simply mimic big stakeholders. - Automation with selective human oversight: AI stewards would handle routine votes automatically and surface only the significant or contentious issues for human review, reducing voter fatigue while preserving agency. - Quality control via prediction markets: To counter spammy or low-quality proposals — a problem amplified by generative AI — Buterin suggests running prediction markets where agents bet on the probability a proposal will pass. Accurate bets would be rewarded, creating financial incentives for useful contributions and penalties for noise. Why it matters The plan is pitched as a way to move away from simple token-based delegation to a model that preserves individual intent, reduces centralization, and scales governance without forcing every participant to become an expert. It also stitches together several cryptographic and privacy-preserving tools already gaining traction in Web3, positioning personal AI stewards as a practical bridge between human values and machine-scale decision-making. The proposal follows Buterin’s recent critique that many DAOs are suffering from low participation and consolidating power — an issue he’s now trying to address with a technical, privacy-centric governance redesign. Read more AI-generated news on: undefined/news