• workshop_dense_100#slide 43

Within RAIDT, a timestamp is the temporal marker attached to a recorded run, showing when the configured use occurred and helping stabilise that run as an inspectable case. The foundations paper includes it directly in the minimal run record alongside a stable run ID, and defines the run-level evidence pack as the scored object from which governance claims are assessed. In that sense, timestamp is not peripheral metadata: it is one of the fields that makes a run reconstructable as a bounded event rather than as a vague recollection of system use.

A simple RAIDT-style example is the illustrative entry 'RUN-2026-01-20-001 / 2026-01-20 10:15'. On its own, the date-time is modest; within the run-level evidence pack, however, it links the run to the prompt template version, model deployment identifier, retrieval snapshot identifiers and hashes, output integrity records, and reviewer checks that were active at that moment. This is especially important because RAIDT treats influence methods as governance interventions: prompt structure, retrieval augmentation, alignment controls, and adapter versions may all change over time, so reviewers need to know which configuration state produced the output.

Timestamp matters in RAIDT because the framework takes run as the unit of governance and expresses governance quality through a score profile across the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability). That score profile is only credible when evidence is temporally pinned to the specific use event being assessed. The timestamp therefore supports reconstruction, sequence analysis, change control, incident investigation, longitudinal comparison, and the practical meaning of the anchors 1=missing / 3=partial / 5=audit-ready. It helps turn logs and traces into accountable organisational evidence.

In the healthcare discharge-summary scenario used in the papers, a hospital records a run ID, timestamp, structured prompt version, model deployment ID, retrieval snapshot from an internal clinical guideline corpus, output hash, and clinician oversight decision. Suppose a patient later questions why follow-up instructions appeared incomplete. The timestamp lets reviewers recover which note version and guidance snapshot were available at that exact run, and whether the disputed summary was generated before or after a template revision.

That matters because the hospital can then score the run-level evidence pack properly rather than relying on staff memory. If the record shows the run happened at 10:15 under the approved structured prompt and reviewed configuration, the organisation can investigate whether the failure arose from the model output, retrieval content, or human review. The timestamp is therefore essential to both explanation and accountability in a high-stakes workflow.

• 08-RAIDT_Foundations_M_V50
• 13-RAIDT-Evidence-Review_M_v10
• 18-RAIDT-Technical-Foundation_M_v04