• qa_deck_100#slide 89 · Standards and policy interoperability

RAIDT needs standards and policy interoperability because the governance problem is no longer the absence of principles; it is the absence of comparable proof. The papers show that the EU AI Act, ISO/IEC 42001, and the NIST AI RMF all expect organisations to demonstrate documentation, oversight, monitoring, accountability, and post-deployment review, yet they express those expectations through different legal and managerial vocabularies. Without an interoperable evidence layer, the same generative AI deployment can face several parallel evidence regimes, producing fragmented files, repeated interpretation work, and duplicated audits. RAIDT addresses that compliance-measurement gap by turning governance claims into inspectable run-level records rather than leaving them as narrative policy statements.

Interoperability in RAIDT therefore means that the same run-level evidence pack can be reused across frameworks without pretending that the frameworks are identical. This matters because generative AI risk materialises in configured use: one prompt, one retrieval context, one output, one review chain. General policies, model cards, or system-level descriptions remain useful, but they do not by themselves reconstruct what happened in a contested run. By treating the run as the unit of governance, RAIDT gives regulators, auditors, buyers, and operational teams a common evidence object and a score profile across the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability). That shared structure reduces repeated re-audits, supports cross-border comparison, and allows legal regimes to remain distinct while still recognising the same underlying governance evidence.

Consider a public-service eligibility workflow that uses generative AI to draft explanations for applicants. The legal team wants evidence for transparency and oversight, procurement wants supplier assurance, and internal audit wants something it can sample later. If each function asks for a different dossier, the organisation creates three overlapping document trails and still may not know what happened in one disputed case.

With RAIDT, the organisation captures one run-level evidence pack for each material case: prompt version, policy clause used, model deployment, retrieved source snapshot, output hash, reviewer step, and follow-up checks. The legal team can map the pack to regulatory duties, procurement can require the same structure from suppliers, and audit can inspect the same run as if it were a transaction. The value of interoperability is therefore concrete: one evidential object supports several governance conversations without forcing them into one legal template.

• 10-RAIDT_Policy_Pathways_M_V50
• 14-RAIDT-Policy-Motivation_M_v11