Aerospace assurance crosswalk

Field-to-source map

These are review prompts. Follow the primary source and the applicable certification or maintenance process before making a real-world decision.

1. EASA · proposal

   Concept of operations

   Card fields: system.intended_use, human_ai_teaming.role

   Reference: DS.AI.100 at EASA

   Applies when: A vendor describing an AI-based aviation system's concept of operations and end-user role.

   Why linked: The fields disclose intended use and the allocation of work between system and human.

   Limit: A filled card does not demonstrate the proposal's objective or an accepted ConOps.

2. EASA · proposal

   AI level classification

   Card fields: assurance.easa_ai_level, human_ai_teaming.role, human_ai_teaming.human_override_available

   Reference: DS.AI.110 / AMC1 AI.110 at EASA

   Applies when: Vendor-declared proposed EASA AI level for the relevant high-level task.

   Why linked: Records the claimed level and the human authority posture for review.

   Limit: Level 3B is reserved in the proposal. Level 2A, 2B and 3A have different human authority; a boolean alone cannot establish correct classification.

3. EASA · proposal

   Operational domain

   Card fields: odd.operating_conditions, odd.excluded_conditions, odd.condition_evidence, odd.coverage_evidence_refs

   Reference: DS.AI.120 / AMC1 AI.120 at EASA

   Applies when: An AI-based system with defined operating conditions.

   Why linked: Declares operating limits and links each declared condition to evidence references.

   Limit: Linked evidence is not verified by this crosswalk; declared-bin coverage is not real-world representativeness or approval.

4. EASA · proposal

   Risk and assurance allocation

   Card fields: assurance.target_dal, assurance.standards_posture

   Reference: DS.AI.130 at EASA

   Applies when: When an applicant has a documented safety assessment and allocated assurance level.

   Why linked: Captures a vendor's declared target and related standards posture for later review.

   Limit: No DAL can be inferred from the AI level alone. The card cannot assign, verify, or approve a DAL.

5. EASA · guidance

   Learning assurance

   Card fields: learning_assurance.process_summary, learning_assurance.evidence_refs, data_management.dataset_manifest_refs

   Reference: Issue 2, learning assurance topic at EASA

   Applies when: Learning-based aviation systems within the concept paper's Level 1 or 2 scope.

   Why linked: Points a reviewer to process and dataset evidence without copying restricted datasets into the card.

   Limit: Concept-paper alignment is not an EASA approval or complete learning-assurance demonstration.

6. FAA · roadmap

   FAA learning system

   Card fields: system.intended_use, learning_assurance.process_summary, data_management.split_policy

   Reference: Guiding principles: learned versus learning AI; work within the aviation ecosystem at FAA

   Applies when: FAA-oriented safety-assurance discussion for AI in aircraft and related systems.

   Why linked: Prompts explicit intended-use and learning-process documentation.

   Limit: The roadmap sets research and policy direction, not a product-specific certification method.

7. EUROCAE · draft

   ED-324 status

   Card fields: assurance.standards_posture

   Reference: ED-324 (draft); ARP6983 (joint work identifier) at EUROCAE

   Applies when: Only where the eventual ML process standard is relevant to the system and applicant.

   Why linked: A status field can disclose that the standard is being monitored or evaluated.

   Limit: No published ED-324 objective or conformity claim is made. Target publication dates can move.

8. SAE International · licensed published

   ARP4754B applicability

   Card fields: assurance.target_dal, assurance.standards_posture

   Reference: ARP4754B; objective ID not verified at SAE International

   Applies when: Aircraft and system development where the applicant identifies an applicable certification basis; not automatically MRO analytics.

   Why linked: Records an identifier-level standards posture and allocated system-level assurance target.

   Limit: Licensed objective text and objective numbers are not reproduced or asserted.

9. RTCA · licensed published

   DO-178C applicability

   Card fields: assurance.standards_posture

   Reference: DO-178C / ED-12C; objective ID not verified at RTCA

   Applies when: Airborne software when it forms part of the applicant's accepted assurance approach; not automatically off-aircraft maintenance tools.

   Why linked: Records that a vendor cites the standard identifier and can link its own evidence.

   Limit: This row does not claim a particular DO-178C objective, software level, or compliance.

10. NIST · voluntary

    NIST AI RMF context

    Card fields: system.intended_use, odd.operating_conditions

    Reference: MAP 1.1 at NIST

    Applies when: Voluntary risk-management context for organizations using or developing AI.

    Why linked: Makes purpose and operating context explicit for human review.

    Limit: NIST AI RMF is voluntary and not an aviation approval checklist.

11. NIST · voluntary

    NIST AI RMF measurement

    Card fields: odd.condition_evidence, data_management.leakage_check_ref

    Reference: MEASURE 1.1 at NIST

    Applies when: Voluntary choice of measurement methods for significant identified AI risks.

    Why linked: Points to bounded ODD evidence and leakage checks to be judged on their merits.

    Limit: A reference or hash does not prove measurement quality, dataset rights, or representativeness.

12. EASA · regulation

    EU maintenance release separation

    Card fields: human_review.staff_id, human_review.authority_ref, signature.signature_b64url

    Reference: 145.A.50 at EASA

    Applies when: EASA Part-145 maintenance where an appropriately authorised certifying staff member issues a separate release to service.

    Why linked: A review card can point to a human and the organisation's separate authorisation record.

    Limit: A review-card signature does not verify staff privileges or issue a certificate of release to service.

13. U.S. Government Publishing Office / FAA · regulation

    U.S. maintenance record separation

    Card fields: finding.summary, human_review.staff_id, human_review.authority_ref

    Reference: 14 CFR § 43.9(a) at U.S. Government Publishing Office / FAA

    Applies when: U.S. maintenance records under this section; other inspection or operator programmes may have different rules.

    Why linked: An AI review card can travel as supporting evidence alongside the controlled maintenance record.

    Limit: It does not replace the required work description, completion date, or qualifying signer's record.