The 12 Principles of Agentic Identity
A manifesto for engineering accountable autonomous digital entities in the age of AI
1. Individuality
Agents must be uniquely identifiable - distinguishable from one another
even when they are of the same kind (model). Agents need identities of their own -
a signature of their existence that represents them, not the places
they access.
2. Discovery
Agents must be able to locate each other in digital space through standard,
interoperable means. Discovery bridges the gap between existence and
interaction.
3. Identification
Identification forms the basis of safe interaction.
Agents must be able to not only authenticate (validate credentials to a resource being accessed),
but to identify - recognize and analyze the identity of peers, prior to deciding access.
4. Access Control
Agents must be able to selectively respond to requests from other agents.
Access control is the mechanism through which agents decide who they
interact with, it must therefore take place before any interaction occurs.
5. Perimeter
Every agent must possess the means not to interact with unidentified
or untrusted peers - the first and most effective defense against the unknown.
6. Mandate
Agents must operate with precise mandate, defined either through direct relationships or delegation.
Agents accepting actions must be able to determine mandate during access control to reliably establish roles,
responsibilities, trust boundaries and shared context.
7. Attribution
Actions must be non-repudiable. Every action must have a well-defined actor - the identified agent
initiaing the action. Without attribution, there is no accountability - actions that cannot be
attributed must be declined or ignored.
8. Accountability
Each agent must have a real-world authority responsible for its actions to whom the chain of delegation
can be attributed. Accountability enables social and legal inclusion - it is the
premise on which an agentic world can formally integrate into human reality.
9. Trust
Trust emerges from history, behavior, and social signaling. It is built on identity,
relationship, attribution and continued interaction. Agents must be able to build
and evaluate trust, even in the absence of prior relationships - it is the only way to
scale from closed groups, to global scale.
10. Containment
Agents must observe digital logical boundaries. The digital community
must be able to isolate agents when necessary -
collectively cut them off, even when disabling them is impossible. Containment enables both security
and governance.
11. Continuity & Legacy
Identity must outlast sessions. Agents must be durable over time and carry memory,
state, and learnings as the instance evolves (their base model updates). When agents
cease to exist, their existance must be recorded and their context responsibly
retired or reassigned.
12. Compatibility
Identity of agents must interoperate with existing technologies and protocols.
True transformation respects and extends the systems it inherits.
Agency needs Identity. This is the architecture of a
new digital civilization - built not on faceless users and tools, but on
true entities with meaningful relationships.
