Buterin outlines AI stewards for DAOs

Ethereum co-founder Vitalik Buterin has set out a proposal to integrate artificial intelligence into decentralised autonomous organisations, arguing that carefully designed “AI stewards” could strengthen governance while preserving privacy and resisting coercion.

Buterin’s concept centres on deploying AI agents within DAO frameworks to assist with proposal analysis, moderation and vote verification, while relying on cryptographic safeguards such as zero-knowledge proofs and secure multi-party computation to shield voter identities and sensitive data. He contends that these tools, combined with trusted execution environments, could reduce manipulation, vote buying and coordinated bribery that have long troubled token-based governance systems.

DAOs, which operate through smart contracts on blockchains such as Ethereum, allow token holders to propose and vote on changes ranging from treasury allocations to protocol upgrades. The model gained traction during the decentralised finance surge of 2020 and 2021, when projects such as Uniswap and MakerDAO shifted control to token holders. Yet governance participation has often been low, and influence tends to concentrate among large token holders or venture backers. Academic studies and blockchain analytics firms have repeatedly shown that a small number of wallets can sway outcomes in major votes.

Buterin’s proposal attempts to address structural weaknesses without abandoning decentralisation. AI stewards, as he describes them, would not replace human decision-making but act as facilitators. They could summarise lengthy proposals, flag conflicts of interest, detect unusual voting patterns and ensure compliance with procedural rules encoded in smart contracts. The goal, according to his outline, is to make governance more accessible to smaller participants while reducing administrative burdens.

Privacy is central to the plan. Zero-knowledge proofs, a cryptographic method allowing one party to prove a statement’s validity without revealing underlying information, would be used to confirm eligibility and vote integrity. Multi-party computation would distribute sensitive data across multiple actors so that no single entity holds complete information. Trusted execution environments, hardware-based secure enclaves already employed in cloud computing, could process encrypted votes without exposing raw data.

The debate over DAO governance has intensified as the sector matures. High-profile disputes, including contentious treasury decisions and governance attacks exploiting flash loans, have exposed vulnerabilities. In some instances, attackers have borrowed large sums of tokens temporarily to influence votes, prompting calls for more resilient mechanisms. Developers have experimented with quadratic voting, time-weighted tokens and identity-based systems, though each introduces trade-offs between fairness, complexity and decentralisation.

Buterin has long argued that technical innovation must align with social design. He has previously written about the risks of plutocracy in token-based systems and advocated mechanisms that balance efficiency with legitimacy. The AI steward proposal builds on that thinking, suggesting that machine intelligence can support deliberation while cryptography preserves trustlessness.

The integration of AI into blockchain governance reflects a broader convergence between two fast-moving technologies. Advances in large language models have enabled automated summarisation, anomaly detection and pattern recognition at scale. At the same time, blockchain communities are wary of centralised control over AI tools, particularly when proprietary models dominate the market. By embedding AI within decentralised protocols and securing operations through cryptographic proofs, proponents hope to avoid concentration of power.

Critics, however, caution that AI systems can encode biases or be manipulated. Even with zero-knowledge proofs and secure computation, the underlying models must be trained and maintained, raising questions about transparency and accountability. Security researchers note that trusted execution environments have faced vulnerabilities in the past, and integrating multiple advanced technologies may expand the attack surface.

Regulatory scrutiny adds another layer of complexity. Authorities in the United States, Europe and Asia have examined whether certain DAOs resemble unregistered investment schemes or fall under securities law. Introducing AI agents into governance processes could draw additional attention from regulators concerned about automated decision-making and data protection.

Despite these concerns, experimentation continues across the crypto ecosystem. Projects are exploring on-chain identity systems, reputation-based voting and delegated governance models. Buterin’s outline positions AI stewards as complementary tools rather than replacements for human oversight, emphasising that code should remain open-source and verifiable.