[Use Case] Cross-jurisdiction trust interoperability for AI-produced health compliance evidence — a machine-readable evidence-signal approach

Complete every field. Replace the angle-bracket placeholders. Tick boxes with [x].

1. Identification

Title: Cross-jurisdiction trust interoperability for AI-produced health compliance evidence — a machine-readable evidence-signal approach
Submitting organization: Independent (submitted in an individual capacity as an independent expert)
Contact name: Steven Zhao (independent expert, ORCID 0009-0001-0512-1237)
Contact email: zhaoxinghua09@gmail.com

Sector (tick all that apply; a cross-industry case may tick more than one):
- [x] Health
- [x] Other (specify): Professional regulatory-compliance services (cross-industry applicability)

2. The situation

Plain language description (3 sentences max): AI agent teams now prepare medical-device regulatory compliance deliverables for manufacturers under multiple national regimes, but the trust signals attached to that machine-produced evidence do not interoperate across jurisdictions or with existing health-data standards. A receiving regulator or trading partner in another country cannot machine-read whether the evidence was produced under a verifiable governance regime. This submission proposes a machine-readable evidence-to-trust-signal mapping that links into existing compliance and health-data pipelines without requiring any change to national regulations.

Actor: Medical-device manufacturer (client); AI agent team (multi-agent delivery); receiving parties (client QA/regulatory staff; foreign regulators; trading partners)
Action: Agents research, draft, verify and assemble regulated compliance deliverables; the deliverables carry structured, verifiable trust signals intended for cross-border reliance
Decision required: Under what conditions may a foreign regulator or trading partner accept an agent-produced health compliance evidence package as trustworthy, given that national regulations and health-data standards differ?
Problem encountered: No shared data model lets one jurisdiction's health-trust signals be read by another; agent-produced evidence lacks a machine-readable form that existing HL7 FHIR / IHE infrastructure can consume
Current mitigation: An operational identity-and-trust lifecycle model with structured evidence semantics; an operational reference deployment together with twelve medical-device compliance tooling packages, exercised in production across NMPA, FDA, EU MDR and JP PMDA submission workflows
Residual gap: Cross-jurisdiction interoperability of trust signals for health evidence is undefined; FG-TIDA has no health-interoperability use case beyond use case #14, and no behavioural-trust-signal data model yet bridges compliance evidence to health-data standards

3. Actors and context

Taxonomy roles involved
- [x] Principal — possesses the authority being exercised; source of the mandate
- [x] Relying party — must decide whether to accept the agent's authority (foreign regulator / trading partner)
- [x] Builder
- [x] Deployer / Owner
- [x] Agent instance
- [x] Infra provider
- [x] Attestor

Mandates in this case

Cross-border?
- [x] Yes — jurisdictions involved: China (NMPA), USA (FDA), EU (MDR), Japan (PMDA)
- [ ] No

Embodied (physical actuation or sensing)?
- [ ] Yes
- [x] No

Agent action type (tick all that apply):
- [ ] Read-only
- [x] Consequential (reversible)
- [x] Irreversible

Crosses organisational boundary?
- [x] Yes (specify): agent team → client organisation → foreign regulator / trading partner (three or more trust boundaries)
- [ ] No

Risk level
- [ ] Low
- [ ] Medium
- [x] High

4. Theme relevance

ToR anchors addressed: A.2.5 (cross-jurisdiction / cross-industry trust interoperability, including healthcare), A.2.7 (trust control plane reference architecture), A.2.8 (behavioural trust-signal data model), A.3.1 (health use case).

5. Proposed contribution — a machine-readable evidence-signal approach

The submission offers a concrete gap-closing approach rather than a publication:

1. Non-destructive adoption. Adopting organizations link our machine-readable trust signals into existing compliance and health-data pipelines (HL7 FHIR Provenance / AuditEvent / Consent / VerificationResult; IHE XUA / ATNA / IUA) without altering any national regulation. This keeps adoption friction low and respects each jurisdiction's sovereign rulemaking.
2. Evidence, not logs. Trust signals are structured, verifiable evidence objects (distinct from audit logs), directly consumable by a behavioural trust-signal data model of the kind ToR A.2.8 envisions.
3. Standards alignment. The approach maps onto ISO/IEC 42001:2023 (AI management system — a certifiable AI management-system standard), ISO/IEC 23894:2023 (AI risk management), ISO/IEC TR 24028:2020 (overview of trustworthiness in AI — a Technical Report, not certifiable), ISO/IEC 22989:2022, ISO/IEC 5723:2022, ISO/IEC 5338:2023, the NIST AI Risk Management Framework, and the OWASP Agentic Top 10 — rather than introducing a parallel standard.
4. Production grounding. The same approach is exercised in production across NMPA, FDA, EU MDR and JP PMDA medical-device compliance workflows, demonstrating that the evidence semantics are applicable in multiple jurisdictions.

Scope note. The model referred to here is an identity, trust and lifecycle model, not an AI-governance framework.

6. Relation to existing work

• Builds on and extends use case #14 (agent-team delivery of regulated medical-device compliance work) from single-jurisdiction production to cross-border trust interoperability — the dimension #14 itself flags as a residual gap.
• Complements silent-failure-catalog work (verifier-side failure semantics) by giving those failure semantics a cross-jurisdiction, health-interoperable transport form.

7. References (real standards, verified)

• HL7 FHIR R4 (v4.0.1, current normative baseline) and R5 (5.0.0): resources Provenance, AuditEvent, Consent, VerificationResult. AuditEvent is defined on IHE-ATNA audit record definitions, originally RFC 3881, now maintained by DICOM.
• IHE IT Infrastructure: XUA (Cross-Enterprise User Assertion, ITI-40, final text), ATNA (Audit Trail and Node Authentication, ITI-19/20, final text), IUA (Internet User Authorization, ITI-71/72/102/103, Trial Implementation).
• ISO/IEC 42001:2023 (AI management system — a certifiable AI management-system standard), ISO/IEC 23894:2023 (AI risk management), ISO/IEC TR 24028:2020 (overview of trustworthiness in AI — a Technical Report, not certifiable), ISO/IEC 22989:2022, ISO/IEC 5723:2022, ISO/IEC 5338:2023.
• EU: Regulation (EU) 2024/1689 (AI Act); Medical Device Regulation 2017/745; In Vitro Diagnostic Regulation 2017/746; GDPR Article 9 (special-category health data) and Article 22 (automated decisions).
• China: Personal Information Protection Law Articles 38-40 (cross-border provision); the 2025 Good Manufacturing Practice for Medical Devices (Announcement No. 107) takes effect in November 2026.
• NIST AI Risk Management Framework; OWASP Agentic Top 10.

The underlying material, including its provenance and licensing, is maintained in the author's own repository: https://github.com/zhaoxinghua09-cell/lgd-theory