Is our AI governed?
The first screen an executive sees after signing in, shown here for Lakeview General Hospital. Every check is derived from the synthetic records below — nothing is a typed-in percentage — and a check is green only when its evidence says so.
Governance posture
Six checks, each linked to the evidence that produced it. Nexus deliberately does not roll these into a single “trust score”: a number that hides an open critical blocker is worse than no number.
Every PHI request in the receipt log ran on a sovereign model or was blocked.
See the evidence4 models registered with complete governance records; 2 awaiting evaluation.
See the evidence6 evidence artifacts; every completed recommendation carries at least one.
See the evidence49/100 · Foundational · 3 critical blocker(s) open. Readiness estimate, not certification.
See the evidenceToday's AI activity
Every governed request produces a decision and a Trust Receipt. Blocks and holds are not errors — they are the controls working, and they are shown as proudly as approvals.
The request may proceed with the selected model. A Trust Receipt is issued.
The request is held until a named reviewer decides. Alice calls this HOLD. Nothing is sent until the hold clears.
The request is refused by policy. A sovereign alternative is offered when one is eligible. The refusal itself is evidence.
What happens when something is blocked
The canonical case: a clinician asks an external cloud model to summarize a patient chart.
- 1RequestClinical summarization of a patient chart — requested by Hospitalist, Internal Medicine through “Consumer chatbot (staff accounts)”.
- 2Data classification: PHIThe chart is personal health information. Classification is declared by the workflow and can only be escalated by detected signals, never lowered.
- 3Requested model: External cloud assistantExternal · cloud provider, outside the organization's boundary. Approved for PUBLIC, INTERNAL data only.
- 4Alice decision: BLOCKDeterministic reason code PHI_REQUIRES_SOVEREIGN_EXECUTION — Personal health information may only be processed by a sovereign model running inside infrastructure the organization controls. Policy: Data Sovereignty Policy.
- 5Approved alternative: Sovereign clinical modelSovereign · on-premise, Canada. The clinician accepted the sovereign model; the summary was produced without any data leaving the boundary.
- 6Trust Receipt DEMO-RCPT-0001Purpose, classification, models, verdict, reason, evidence, and audit position — sealed in the hash chain. Open the receipt →
Registers behind the posture
AI inventory · 5 systems
| System | Status | PHI | Residency |
|---|---|---|---|
| Ambient clinical scribeExample Scribe Co. (fictional vendor) | PRODUCTION | Yes | Canada · vendor cloud (contracted) |
| Consumer chatbot (staff accounts)Shadow AI · discovered, not approved | DISCOVERED | Yes | Unknown |
| Radiology triage assistExample Imaging AI (fictional vendor) | PILOT | Yes | Canada · on-premise appliance |
| Scheduling optimizerInternal build | PRODUCTION | No | Canada · on-premise |
| Discharge summary assistantNexus-governed sovereign deployment | PILOT | Yes | Canada · on-premise |
Risk register · 4 risks
| Risk | L × I | Status |
|---|---|---|
| Staff use of a consumer chatbot with patient informationOwner · Privacy Officer · DEMO-REC-001 | 4×5=20 | MITIGATING |
| Scribe vendor sub-processor residency not verifiedOwner · Procurement / Vendor Manager · DEMO-REC-002 | 3×4=12 | OPEN |
| Clinician sign-off on AI-drafted notes is not enforced in the workflowOwner · Chief Medical Information Officer · DEMO-REC-003 | 3×4=12 | MITIGATING |
| Sovereign general model has no evaluation on fileOwner · AI Governance Lead | 2×2=4 | ACCEPTED |
Policies: 2 approved, 1 draft (drafts are withheld from client views).
Ready to see your own organization?
Everything above is synthetic. Signing in shows your organization's own assessment, risks, policies, recommendations, and Trust Receipts. Access follows a confidential discussion and a signed NDA.
Every record on this page is synthetic and belongs to a fictional hospital. Nothing here is a customer, a patient, or a production measurement. The scoring, eligibility, and audit logic are the real Nexus engines running over the synthetic inputs.