• qa_deck_100#slide 38 · Minimum run identity and configuration record

In RAIDT, a timestamp is part of governance because it helps constitute the run-level evidence pack for one configured use of generative AI in context, rather than serving as a clerical filing detail. The papers repeatedly define a run as occurring at a specific time, and they treat run as the unit of governance because accountability, contestability, and review attach to what happened in that concrete use event. A timestamp therefore identifies when the prompt version, model deployment, retrieval snapshot, alignment policy, checks, and oversight decision were actually in force. Without that temporal anchor, an organisation cannot reliably reconstruct which controls applied, or show whether a disputed output was produced under approved conditions.

This is why timestamping belongs to governance rather than administration. Administrative metadata may help organise records, but RAIDT uses temporal evidence to support auditability, traceability, post-incident reconstruction, audit sampling, and learning across runs. The foundations paper explicitly links timestamps with hashes and access logs to support chain-of-custody in audits and incident investigations, while the technical foundation argues that provenance turns governance from memory into evidence. In practice, the timestamp helps make the score profile defensible across the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability). If the time marker is absent, reviewers cannot confidently relate a run to the correct versions, surrounding events, or later incident reports, so evidence quality falls towards anchors 1=missing / 3=partial / 5=audit-ready rather than reaching robust governance readiness.

Consider a public-service eligibility workflow in which a council officer uses a GenAI assistant to draft advice from an internal policy corpus. A citizen later challenges the advice. The timestamp in the run-level evidence pack is governance-relevant because it shows exactly when that run occurred, which policy version was retrieved, which prompt template was active, and which reviewer approved or amended the output.

If the policy changed a week later, the timestamp distinguishes the contested run from later configurations and allows auditors to reconstruct the sequence of retrieval, generation, and approval. It also lets the organisation connect the run to the right incident ticket and review log. Without the timestamp, the record is only loosely administrative; with it, the run becomes a reviewable governance object that can be contested and defended.

• 08-RAIDT_Foundations_M_V50
• 18-RAIDT-Technical-Foundation_M_v04