TechFlow News: On February 23, according to Cointelegraph, Vitalik posted on X (formerly Twitter) this past Sunday proposing the integration of personal AI large language models (LLMs) into decentralized autonomous organization (DAO) governance to address the long-standing issue of low participation rates. Currently, DAOs average only 15%–25% voter turnout—a situation that not only risks centralizing power but also exposes governance to potential attacks.
Vitalik pointed out that the core challenge facing democratic and decentralized governance is the “scarcity of human attention.” Existing delegation mechanisms concentrate decision-making authority in the hands of a few representatives, effectively silencing most members’ voices. He proposed that personal AI assistants could infer users’ preferences from their writing history, conversation logs, and explicit statements—and vote on their behalf. When uncertain about an important issue, the AI should proactively query the user and provide relevant background information.
Regarding privacy protection, Vitalik suggested running personal LLMs within a “black-box” environment to process sensitive data—outputting only final voting decisions—thus enabling governance decisions involving confidential information while safeguarding user privacy.
