Privacy protocols for continuous delegation at trillion-agent scale

Determine protocols and standards that ensure robust data minimization and protection of user data across billions of autonomous AI agents operating under continuous delegation within the Internet of AI Agents, including mechanisms for privacy preservation and compliance at planetary scale.

Background

The paper argues that autonomous AI agents will operate continuously, delegate tasks to helper agents, and cross organizational boundaries, creating novel privacy risks not addressed by existing web infrastructure. Continuous delegation and agent-to-agent negotiation amplify exposure of user data across numerous actors and platforms.

The authors highlight that current standards provide no clear blueprint for data minimization and privacy at the scale of billions of agents. They mention candidate techniques such as zero-knowledge capability proofs, differential privacy, and unlinkable yet accountable identifiers but emphasize that the protocols to operationalize these remain undefined.