The Human Root of Trust: Establishing Accountability for Autonomous AI Agents

The Human Root of Trust framework addresses a fundamental issue in digital systems: the assumption that a human is always present at the other end. With autonomous AI agents now performing tasks once attributed only to humans, such as managing transactions and signing contracts, there is a pressing need for systems that can attribute actions to accountable humans.

This framework introduces three core pillars essential for establishing accountability in AI systems:

• Proof of Humanity: Ensures that there is a clear association between the agent's actions and a real human.
• Hardware-rooted Device Identity: Establishes device integrity and authenticity, ensuring that actions can be traced back to an identified hardware source.
• Action Attestation: Provides verifiable evidence that actions taken by AI agents are authentic and authorized by a human principal.

The architecture includes a six-step trust chain connecting a human principal to a cryptographic receipt, ensuring thorough traceability of actions. The Human Root of Trust is not a product or a standard but a public domain principle designed to build systems that cryptographically manage and verify accountability.

Implementers, like security engineers, cryptographers, and legal experts, are encouraged to develop and refine the framework, which is freely available without patent claims or user attribution requirements. As AI agents become increasingly prevalent, frameworks like this will be crucial in answering regulators' accountability questions.