AI agent identity: what a Digital Passport is
AI agents already browse, call APIs, write and run code, and chain multi-step tasks, but nothing verifies who built an agent or what it declared it would do. A Digital Passport is a proposed design for that missing layer: a real-time handshake that checks an agent's origin and declared objectives at the moment it acts. This guide summarizes the proposal from Alex London's essay.
What does the Digital Passport check?
The handshake runs when an agent tries to touch a network, platform or service, closer to passport control at a border than to a document filed months earlier. The receiving system checks four things on the spot:
- Who originated this agent?
- What objectives has it declared?
- Does the current action match one of those objectives?
- Is the passport still valid and unrevoked?
Pass all four and the request goes through. Fail one and the system blocks it and logs what happened.
What is the objective registry?
Instead of a free-text mission statement, operators pick from a checklist of pre-defined, legal objectives at registration:
- Web search and summarization
- API calls to listed third-party services
- Form submission on behalf of a user
- File read and write on the user's own device
- Communication, such as email or messaging, on behalf of a user
- Code execution inside a sandbox
- Calendar and scheduling actions
- Financial transactions within a declared limit
A checklist can be checked by machines without a human reading anything, audited later by comparing declared objectives with logged behavior, and it stops an agent from rephrasing its way into a wider mandate.
Does it identify the person running the agent?
No. The handshake verifies the agent, not the person running it. No user identity and no user IP address crosses that layer. It answers one question: is this agent who it claims to be, doing what it said it would do?
How do risk tiers work?
| Tier | Examples | What it takes |
|---|---|---|
| Standard | Search, summarization, read-only API calls | A registered originator and a valid process ID |
| Elevated | Form submission, messaging, scheduling | Stronger originator verification and a declared scope limit |
| Restricted | Financial transactions, sensitive systems, autonomous code deployment | Review by the issuing authority before going live, and a lower threshold for the kill switch |
Friction scales with what the agent can do, not with the size of whoever built it.
What happens when an agent drifts?
The proposal calls for a kill switch, not a punishment: suspend the credential, stop the handshake from clearing, and log the incident for review. It is quarantine for the specific behavior, with no lawsuit needed to respond.
Does it need new technology?
No. The handshake runs over TLS, credentials are signed with ordinary public-key infrastructure, and verification is stateless on the receiving end. The essay argues the registry protocol should be an open public standard, minimal enough for a solo developer to use the same way a large company does.
Frequently asked questions
What is a Digital Passport for AI agents?
A proposed real-time handshake that checks an agent's origin, declared objectives, and credential validity at the moment it acts.
Is the Digital Passport a shipped product?
No. It is a design proposed in Alex London's essay and summarized in this guide.
Does the Digital Passport identify users?
No. It verifies the agent, not the person running it, and no user identity or IP address crosses the handshake.
Why use a checklist of objectives?
Machines can check it automatically, regulators and platforms can audit declared objectives against logged behavior, and it prevents an agent from rephrasing its way into a wider mandate.
Build on it
Open the studio, compare plans, or request enterprise access.