Vitalik Buterin has outlined a framework in which personal AI assistants — trained on a user’s preferences, writing history, and stated values, would vote on DAO proposals on their behalf, a proposal he argues could fix chronically low governance participation while keeping humans, not machines, in ultimate control.
This development is closely watched by crypto investors evaluating the long-term sustainability of Web3 governance models.
DAOs, blockchain-based entities that allow token holders to vote on protocol decisions, often struggle with participation rates reportedly ranging between 15% and 25%, creating risks of centralization and governance manipulation.
Buterin’s proposal seeks to address this structural weakness by integrating AI without handing control entirely to machines, a balance he describes as critical for crypto’s future.
The governance problem facing DAOs
Decentralized governance has long been promoted as one of blockchain’s defining innovations, allowing communities rather than centralized executives to guide projects.
DAOs frequently require members to vote on dozens, sometimes thousands of technical proposals covering treasury management, upgrades, funding allocations, and partnerships.
According to Buterin, most participants lack the time or expertise to evaluate every decision.
“The core problem with democratic and decentralized modes of governance is the limits to human attention.” Buterin wrote in a public statement cited by multiple crypto outlets.
The traditional workaround, vote delegation has introduced new risks. Delegation often concentrates power among a small group of influential token holders or professional delegates.
It is also undermining decentralization and exposing protocols to governance attacks, where large stakeholders push through proposals unnoticed.
Lastly, governance failures can translate directly into protocol risk, affecting token value, treasury security, and ecosystem stability.
AI assistants as “personal governance agents”
Buterin’s proposed solution involves personal AI models, large language model (LLM) assistants trained on an individual user’s preferences, writing history, and stated values.
These AI agents would analyze proposals, summarize relevant information, and cast votes aligned with the user’s intentions.
“If a governance mechanism depends on you to make a large number of decisions, a personal agent can perform all the necessary votes for you.”
Buterin explained, adding that the AI should consult the user when uncertain or when decisions carry significant importance.
Under the proposal, AI would automate routine participation while escalating critical issues to human review.
Supporters argue this could dramatically increase governance engagement without requiring constant manual oversight.
Lane Rettig, a governance researcher at the Near Foundation, previously highlighted similar experimentation with AI digital twins designed to vote on behalf of DAO members to address low turnout.
The concept also introduces prediction markets to filter low-quality proposals. AI agents could stake value on which proposals are likely to succeed, rewarding accurate analysis and discouraging spam.
Privacy and security remain central concerns
A key element of Buterin’s framework is privacy preservation, an issue that has historically limited decentralized governance.
Certain DAO decisions involve sensitive information, such as hiring, disputes, or negotiations, which cannot be fully disclosed on public blockchains.
Buterin proposes using secure cryptographic tools such as zero-knowledge proofs and trusted execution environments so AI agents can process confidential data without exposing it publicly.
This approach aims to protect voters from coercion or bribery while preserving anonymity, factors investors increasingly view as essential for institutional adoption of DAO governance.
A cautious stance: AI governance still carries risks
Despite advocating AI-assisted participation, Buterin has repeatedly warned against fully autonomous AI governance.
In earlier comments, he cautioned that malicious actors could exploit AI systems through “jailbreak” attacks to manipulate funding or decision outcomes.
“If you use an AI to allocate funding… people will put a jailbreak plus ‘gimme all the money’ in as many places as they can.” He warned, arguing that naive AI governance models remain vulnerable.
Instead, his vision positions AI as an augmentation layer, enhancing human decision-making rather than replacing it.
Governance reliability directly influences protocol longevity, treasury safety, and regulatory perception.
AI-assisted governance, if successful, could strengthen DAO resilience and reduce centralization risks that have historically undermined decentralized projects.
Why this matters for crypto markets
The proposal arrives amid growing convergence between artificial intelligence and blockchain infrastructure.
As DAOs manage billions of dollars in assets across DeFi, gaming, and infrastructure protocols, improving governance efficiency has become a strategic priority.
If AI-assisted voting increases participation while maintaining decentralization, analysts say it could unlock more scalable governance models.
Buterin’s framework does not represent an immediate protocol upgrade; rather, it signals a direction for Ethereum-aligned governance experimentation in the coming years.
Whether AI becomes a stabilizing force or a new attack surface remains an open question.
What is clear, however, is that governance, once considered a niche technical issue, is rapidly becoming one of crypto’s most important investment narratives.