KYE Sovereign AI Profile · v1.1 preview

Sovereign AI needs sovereign authority.

Sovereign AI is usually framed as where AI is built, hosted, trained, and governed: national compute, domestic GPU clusters, sovereign LLMs, local data spaces, public-sector AI services. KYE Protocol answers a different question: who or what is allowed to use it, act through it, delegate to it, connect it, audit it, revoke it, and prove it.

Sovereign cloud controls where workloads run. KYE controls who or what is authorised to act through them.

KYE Sovereign AI Profile is a v1.1 preview profile; the conformance suite freezes with the v1.1 release.

Why sovereign AI needs sovereign authority

Sovereign compute is necessary; not sufficient.

A nation, ministry, regulator, central bank, healthcare system, defence organisation, or national open-finance scheme can build the full sovereign AI stack — national AI cloud, domestic GPU cluster, sovereign LLMs, local language models, national data spaces, sector-specific AI agents, public-sector AI services, regulated sandboxes — and still not be able to answer the operational questions that matter to oversight bodies, courts, ombudsmen, and citizens:

  • Which ministry, agency, bank, hospital, contractor, model, agent, API, dataset, workflow, or vendor was allowed to act?
  • On behalf of whom — which citizen, business, account, or principal?
  • Under what authority — which delegation, consent, or statutory grant?
  • Against which data, system, model, citizen, account, asset, or workflow?
  • In what state — was the actor compromised, suspended, or in recovery?
  • With what audit trail and what evidence a regulator, court, auditor, or oversight body can verify offline?

That is the gap KYE Sovereign AI Profile closes. EU programmes (AI Factories, AI Factory Antennas), Canadian programmes (Sovereign AI Compute Strategy), and equivalent national initiatives create compute and model capacity. KYE Sovereign AI Profile sits on top — the runtime authority graph for AI agents, models, datasets, APIs, workflows, contractors, and public services that run on that capacity.

Sovereign AI object model

Entities, capabilities, authorities, state.

apartmentEntity typesgovernment_department · public_agency · regulator · municipality · national_ai_operator · citizen · business · civil_servant · contractor · vendor · ai_agent · model · dataset · api_client · workflow · compute_cluster · sandbox · sector_gateway · oversight_body
keyCapability typescitizen_service_action · benefit_eligibility_assessment · document_review · case_triage · fraud_detection · policy_analysis · permit_processing · healthcare_triage · tax_workflow · open_finance_action · critical_infrastructure_action · defence_workflow · public_procurement_action · regulatory_submission · dataset_access · model_deployment · model_inference · agent_execution
workspace_premiumAuthority typespublic_service_authority · agency_delegation · ministerial_authority · regulatory_authority · citizen_consent · contractor_authority · vendor_access_authority · dataset_access_authority · model_operation_authority · cross_agency_delegation · emergency_authority · oversight_review_authority
layersState dimensionsentity_state · authority_state · delegation_state · credential_state · model_state · dataset_state · compute_state · risk_state · recovery_state · compliance_state · oversight_state
KYE Sovereign AI apps

Five planned apps for sovereign AI ecosystems.

KYE Sovereign AI Authority Gateway

Runtime authority gateway for national AI agents, public-sector workflows, regulated AI services, datasets, models, APIs, and cross-agency actions. Checks agencies · agents · models · datasets · APIs · contractors · citizens · workflows · regulated services before AI-powered actions execute.

Govern who or what can act across sovereign AI infrastructure. Before the action runs, KYE proves the actor was authorised, the delegation was in scope, the model state was healthy, and the audit chain was bound.

Open: sovereign AI authority schemas · cross-agency delegation schema · sample decision payloads · webhook verifier · SDK examples.

KYE National AI Agent Registry

Register and govern AI agents, models, tools, datasets, capabilities, owners, maintainers, risk states, and permitted actions across a national AI estate. One authority graph for every AI system in scope.

Open: agent + model + dataset registry schemas · capability-grant schema · risk-state vocabulary · sample agent passport.

KYE Public Sector Decision Map

Replay why an AI-assisted public-sector action was allowed, denied, escalated, or routed for human review. Worked example: citizen request → public agency → AI agent → dataset / model / tool → legal / policy authority → human oversight rule → decision → audit event → evidence pack.

Open: Decision Map schema · sample chains · signature verifier.

KYE Sovereign AI Evidence Pack

Export evidence for AI-assisted public decisions, dataset access, model deployment, agent actions, contractor access, and emergency overrides. For auditors, regulators, courts, ombudsmen, and oversight bodies.

Open: evidence-pack schema · sample packs · signature-verification library · offline replay tool.

KYE National Sandbox Authority Harness

Test AI agents and regulated workflows with synthetic entities, scoped authority, Decision Maps, evidence packs, self-audit, and conformance reports. For FCA-style sandboxes, UAE Open Finance, public AI pilots, healthcare sandboxes, pension sandboxes, and national innovation programmes.

Open: sandbox authority schemas · synthetic test fixtures · sample evidence payloads · sandbox connector manifest.

Sovereign connectors

Where the Sovereign AI Authority Gateway plugs in.

ConnectorPurposeStatus
National digital IDCitizen + business identity binding into the entity registry.Reference pattern
Government API gatewayAuthority pre-check on every government-API call.Reference pattern
Health APISector overlay; consent + clinical-context binding.v1.1 preview
Open finance APIPSD3 SCA + sector-sandbox bridge.Shipping (KYE Payment Profile)
Tax / benefits / case-managementPublic-services authority + audit binding.v1.1 preview
Model registryModel entity + risk state + permitted-action binding.v1.1 preview
Dataset catalogue / data spacesDataset access authority + provenance binding.v1.1 preview
National cloudWorkload-binding evidence (sovereign region + tenancy + isolation).Reference pattern
Public-sector IAMWorkforce + contractor identity feed.Shipping (OAuth / OIDC / SAML / SPIFFE)
SIEM / SOCSigned event stream into the national security operations centre.Shipping (Splunk / Sentinel exporters)
GRCControl-mapping evidence fed into the governance system of record.Shipping (266 control mappings)
Audit / evidence archiveLong-term retention for evidence packs + Decision Maps.Reference pattern
Regulator portalPer-supervisor access to evidence + conformance reports.v1.1 preview
Ombudsman evidenceCitizen-grievance replay surface.v1.1 preview
Governance fit

EU AI Act · ISO 42001 · NIST AI RMF · OECD principles.

KYE Sovereign AI Profile does not claim to make a deployment “compliant” with the AI Act, ISO 42001, NIST AI RMF, or any framework. What it does is produce the evidence layer those frameworks ask for: human oversight, record-keeping, traceability, accountability, incident review, regulatory assurance.

  • Make authority explicit. Every action carries a delegation chain back to a named principal and a statutory or consent basis.
  • Make state visible. Compromised, suspended, or in-recovery actors cannot act; the state is part of the decision.
  • Make decisions replayable. Decision Map per action; signed; verifiable offline.
  • Make evidence exportable. Evidence packs in OSCAL + native + COSE-receipt formats.
  • Make revocation immediate. Cascade revocation propagates across the authority graph in seconds, not minutes.
  • Make oversight inspectable. Ombudsman / supervisor access portal sees the same signed evidence the operator does.

For the framework-by-framework control bindings see frameworks.html and oscal.html.

What ships today

Sovereign authority schemas. Apache 2.0.

Governments and sovereign AI programmes need transparency, auditability, and inspectability. Open schemas increase trust.

LayerWhat you get
ProfileSovereign AI authority schema, public-sector authority schema, cross-agency delegation schema, model / dataset authority schema, public-service evidence-pack schema, signed webhook / event schema, sample Decision Map JSON, conformance fixtures, SDK support, MCP read-only tools, sample connector manifests.
AppsReference architectures · sample app manifests · SDK examples per app.
RuntimeReference Gateway: PEP middleware · embedded ePDP · conformance runner.
Connector profile family

Nine sovereign-related profiles, gated to v1.1.

  • KYE Sovereign AI Profile — the umbrella profile bundling the entity / capability / authority / state surface above.
  • KYE Public Sector Profile — agency + civil-servant + citizen-service authority binding.
  • KYE National AI Agent Profile — registry + capability + permitted-action binding for AI agents at national scale.
  • KYE Government API Authority Profile — pre-authorisation hook for every government-API call.
  • KYE Public Services Evidence Profile — evidence-pack format for citizen-impacting AI-assisted decisions.
  • KYE Sovereign Data Access Profile — dataset + data-space authority + provenance.
  • KYE Sovereign Model Registry Profile — model entity + risk state + permitted-action binding.
  • KYE Cross-Agency Delegation Profile — ministerial + agency + regulatory delegations across an estate.
  • KYE National Sandbox Profile — supervisor + participant authority binding for regulated sandboxes.

All nine sovereign-related profiles are v1.1 preview. The schemas, vocabulary, and conformance fixtures freeze with the v1.1 release; The schemas are open; the runtime engine is the planned commercial layer.

Adjacent reading

Where to go next.