Signed agent mandates, explained.

A signed agent mandate is a cryptographically signed, scoped and revocable authorization that proves a person granted an AI agent specific permission to act on their behalf. It answers the question every counterparty will soon ask an autonomous agent: "Who said you could do this, and within what limits?"

Think of it as a verifiable power of attorney for software: machine-readable, narrow in scope, time-bound, and checkable by the other side without trusting the agent's own claims.

AI agents are starting to take actions, booking, buying, negotiating, filing, but the systems they interact with have no reliable way to know what a given agent is actually allowed to do. Today that authority is implicit: an API key, a logged-in session, a stored password. Those grant broad, hard-to-audit access and are easy to misuse if an agent is compromised or simply behaves unexpectedly.

A mandate makes authority explicit and provable: the agent presents a credential that states exactly what it may do, and the counterparty verifies it before acting.

A well-formed mandate is built to be narrow and accountable:

• Principal and agent: who is granting authority and which agent holds it, bound to cryptographic keys.
• Scope: the specific actions permitted, for example "book travel up to €500" rather than "act as me".
• Constraints: limits on amount, time window, counterparties or frequency.
• Revocability: the principal can withdraw the mandate, and verifiers can check it has not been revoked.
• Auditability: every use can be logged against the mandate, so there is a record of what the agent did and under whose authority.

Signed agent mandates are an application of self-sovereign identity. They are verifiable credentials whose subject is an action, not a person: the principal uses an identity they control to issue a scoped credential to their agent. Without self-sovereign identity as the foundation, there is no trustworthy place for that authority to originate.

Signed agent mandates are part of validant.ai's research direction, not a shipped feature. They express the third pillar of our mission, personal AI agents that act only in your interest, and they sit on the roadmap alongside self-sovereign identity. Our delivered work today is in AI fairness and digital trust. We describe mandates here because where AI is heading, accountable agency will matter as much as accurate models.