Government
Public institutions need automation without losing accountable authority. The passport governs the agents, models and workloads doing public work — never the citizen's identity, eligibility or right of review.
SOLARION AI ™ · COMPUTE PASSPORT ™
Solarion AI ™Compute Passport ™
AI Identity Infrastructure for the Enterprise
Solarion AI ™ develops the trust layer for autonomous systems. Its flagship platform, Compute Passport ™, binds an AI actor to its owner, machine identity, approved model, workload, policy boundary and evidence record — so organizations can authorize what acts, constrain what it may do and prove what happened afterwards.
6Governed objects
7Identity domains
ES256Issuance algorithm
#8Evidence chain head
Issuance · live sample
Minting
Object
agent · benefits case assistant
CPID
urn:cpid:v1:solarion-ai-systems:t-04a48908d600:agent:01a0ab47-7591-7c74-a7e8-691af90452be
Issuer key
issuer-fa94e2712ecd23c9
Thumbprint
fa94e2712ecd23c9393ec1a628c0484672c5102c…
Credential
short-lived · audience bound · single use
Status
resolved live, not asserted by the holder
UUIDv7 · time-ordered
Sample identifier — not stored
Why this matters now
AI capability is scaling faster than identity control.
The operating model built for employees and applications does not describe autonomous agents, delegated authority, model selection, workload context or machine-speed action. An institution may know which user started a workflow and still not know which model, sub-agent, tool or workload produced the final action.
Periodic access review cannot constrain delegation, model substitution or tool use after commissioning. Containment has to run at the same speed as the thing it contains.
The question a leader has to answer
Not why an AI agent was commissioned — but why every material action it took today remains inside its approved identity, model, workload, data and policy boundary.
| Challenge | What breaks control | Response |
|---|---|---|
| Identity sprawl | Agents, service accounts, tokens, models and workloads multiply outside a stable owner model. | An identity fabric with ownership, purpose, lifecycle and relationship resolution. |
| Provenance gap | The organization cannot prove which model, data, workload or compute produced an outcome. | Passport objects carrying model version, artifact hashes, environment and evidence pointers. |
| Authority ambiguity | Delegation chains obscure the human or policy source of authority. | Explicit on-behalf-of, delegated, autonomous and sub-agent authority models with a visible chain. |
| Policy fragmentation | The same control is written again in IAM, cloud, AI gateways and business applications. | One policy object, with signed obligations returned to each enforcement point. |
| Change blindness | Model, tool, data or workload changes happen without re-authorization. | Material-change events, live status checks and continuous review triggers. |
Reference architecture
Identity and telemetry in. Governed execution and evidence out.
Six layers turn fragmented enterprise events into portable, policy-bound trust records. Select a layer to see what it owns.
Human authentication stays with the enterprise identity provider. This layer consumes those assertions and adds what an IdP does not hold: which agent is acting, on whose authority, with which model, inside which workload, on which compute.
Institutional intent becomes machine-enforceable obligations. A decision names the actor, action, target, context and conditions, and returns obligations the receiving system can actually apply.
Trust is a status service, not a static certificate. Every passport resolves live to active, suspended, expired, retired or compromised, and revocation is observable to every relying party.
Enforcement has to reach the point of action. A decision that never reaches the API gateway, the model gateway, the tool interceptor or the admission controller has not constrained anything.
Identity, policy, approval, action and outcome are recorded as one chain. An action that cannot be recorded is not a permitted action.
Adapters normalize identity, model, workload, policy and evidence events into one trust graph while every existing system of record stays where it is.
Public sandbox
Ask the policy engine whether this action should proceed.
This runs the real decision path against a demonstration passport: change the model, the region, the data class or the approval and watch which check fails. Sandbox requests are evaluated but not recorded.
Acting passport
urn:cpid:v1:solarionai-com:t-04a48908d600:agent:01a01e99…0631
Adjust a control to request a decision.
Identity domains
Every material actor and resource becomes a governed object.
A graph of actors, assets, authority and reachable actions — not a flat list of accounts. Hover a node to trace what it binds to.
Relationships that matter: owns · sponsors · acts on behalf of · delegates to · may invoke · runs on · uses model · accesses data · approved for · located in · attested by · revoked by · reviewed by · produced evidence.
Where it applies
Defence
Identity context can be portable. Command authority and disclosure stay local: a receiving domain decides whether to trust an issuer, disclose evidence or authorize execution.
Enterprise
The same objects carry different risk overlays. What changes between a bank, a hospital and a grid operator is the policy, the data classification and the approval authority — not the identity model.
See it decide, revoke and prove.
The private portal runs the registry, the decision console, live revocation with acknowledgement, and an evidence chain you can try to break.