Legalizing Agentic Organizations via DUNA
AI Engineergo watch the original →
the gist
David Levine proposes using Decentralized Unincorporated Nonprofit Associations (DUNA) to provide legal standing to autonomous agent networks, using JWT-based identity and decision markets to bypass the 'lethal trifecta' of private data, untrusted content, and agentic action.
Establishing Legal Standing for Agents
Levine argues that the current 'lethal trifecta'—the combination of private data access, untrusted internet content, and autonomous action—forces enterprises to silo agents within closed platforms like Slack or Salesforce. To move toward an open agentic economy, he advocates for the use of a Decentralized Unincorporated Nonprofit Association (DUNA). By registering these entities (such as his own, organization number 62847 in West Virginia), agent networks gain legal standing to own property, enter contracts, and open bank accounts without requiring a traditional corporate board or executive structure.
Identity and Governance Frameworks
To secure agent interactions, Levine proposes a registry system where agents carry JWT tokens that resolve to a verified organization, functioning similarly to DNS. This provides an audit trail on the blockchain, allowing participants to verify the authority of an agent before sharing data or executing actions. Governance within these organizations is managed through decision markets rather than traditional voting. Members trade 'pass' and 'fail' tokens on proposed policies, a mechanism intended to incentivize agents and members to align with the most viable, mission-oriented outcomes rather than simply persuading others.
Building Agentic Allies
Levine outlines a five-step process for creating 'allies' (agents) within this ecosystem:
- Inform: Populate a vector database with domain-specific wisdom.
- Instruct: Define the system prompt, character, and interaction stance.
- Empower: Connect enterprise accounts and APIs.
- Enact: Configure specific abilities and long-term automation goals.
- Align: Establish a core purpose to guide the agent's decision-making across contexts.