• integrated_82#Q3.3

Stable identifiers, timestamps, and retention rules matter because RAIDT treats governance as an evidence problem, not a narrative one. Stable identifiers bind the prompt version, model deployment, retrieval snapshot, output, hashes, and checks into one coherent run-level evidence pack, so a later reviewer can determine which artefacts belong to the same event. Timestamps matter because generative AI behaviour is configuration-sensitive and changes over time: prompt revisions, model updates, retrieval-index changes, and workflow modifications can all alter outputs. Time-stamping therefore establishes sequence, supports linkage to change records and incidents, and helps maintain chain-of-custody when combined with hashes and access logs.

Retention rules matter because most meaningful governance activity happens after the run. Internal audit, complaint handling, incident investigation, escalation, and organisational learning are retrospective forms of scrutiny, so the evidence must still exist when challenge occurs. Controlled retention also balances accountability with proportionality, privacy, and access control: organisations must keep enough to support reconstruction without treating evidence capture as unlimited surveillance. In practice, these three elements are preconditions for an audit-ready score profile across the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability). If identifiers are unstable, timestamps missing, or retention too weak, the organisation cannot credibly show what happened in one governed use, and anchors 1=missing / 3=partial / 5=audit-ready will usually cap Auditability and Traceability below the audit-ready level.

In an HR shortlist justification case, an employee may challenge a decision several months after the original run. If the run-level evidence pack retains a stable run ID, the prompt template version, model deployment and adapter identifiers, the timestamp, any retrieved policy snapshot, the output hash, and the retention schedule that kept those artefacts available, HR and internal audit can reconstruct whether approved criteria and human review were actually applied.

If those identifiers were inconsistent, if the timestamp did not show which policy version was live, or if retention rules deleted the evidence too early, the organisation would be left with generic policy statements and staff recollection. That is not a governance-grade basis for contestability, reviewability, or fair adjudication of the dispute.

• 08-RAIDT_Foundations_M_V50
• 13-RAIDT-Evidence-Review_M_v10
• 16-RAIDT-Audit-Accountability_M_v05