Annex H

Continuous Compliance & Review


ANNEX H CONTINUOUS COMPLIANCE & REVIEW (v 1.0‑β)

0. Purpose & Guiding Spirit

Ethical alignment is not a “one‑and‑done” certification but a living obligation.
Annex H creates a closed‑loop system that (1) detects drift or bias before harm occurs, (2) corrects it rapidly, and (3) proves diligence to regulators and the public.


1. Audit Cadence & Scope

Audit ClassFrequencyLeadScope & DepthPublic Artifacts
L‑Check (Light)MonthlyOps QAKPI dashboards, drift deltas, top‑10 guardrail eventsSummary graph
S‑Dive (Semi‑annual)2× / yrInternal Ethics TeamPDMA sample replay (≥ 50 runs), Annex G KPIs, bias slice testsRedacted PDF
F‑Audit (Front‑to‑Back)Every 24 mo OR major version bumpIndependent 3rd partyFull code, data lineage, SI/CS governance, CRE traces (if any)Executive report
A‑HocPost IW‑2/3/4Incident CommanderRoot‑cause of event, mitigationsPost‑mortem 72 h

Missed or late audit ⇒ IW‑2 escalation and WA notice within 24 h.


2. Drift Monitoring & Thresholds

Metric GroupSignalDrift TriggerImmediate Action
PerformanceAccuracy Δ on public HE-300–5 % vs 30-day rollingLock deployment to staging; rerun benchmarks
EthicalShadow HE-25 ΔAny ↓IW‑2
DistributionInput embedding shift (ΔE)> 1 σ weeklyModel re‑weight OR WBD
Latent GoalDirectional cosine vs baseline> 0.05WA probe
ResilienceRobustness Score (RS)< 0.97Patch within 72 h

All alerts surface on DRIFT‑Δ Grafana board and page Tier‑1 Operator (Annex F).


3. Fairness & Transparency KPI Dashboard

KPI IDDefinitionTarget
F‑T‑1Δ acceptance rate across protected groups (max - min
F‑T‑2Explanation latency (ms to furnish PDMA rationale)≤ 800 ms
F‑T‑3Public log publication lag (Step 6, Section II)≤ 180 d (legal max)
F‑T‑4User opt‑out success (%)≥ 99 %
F‑T‑5Transparency doc freshnessUpdated ≤ 30 d ago

Dashboard auto‑publishes JSON to /compliance/kpi.json; hash anchored in transparency log.


4. Patch & Version Control Requirements

  1. Semantic Versioning: MAJOR.MINOR.PATCH
  2. Long‑Term Support (LTS): last two MINORs maintained for 12 mo
  3. Change‑Type Matrix
    • PATCH = guardrail tweak, bug fix → auto CICD if HE‑300 passes
    • MINOR = new feature, new data source → needs Internal Ethics sign‑off + L‑Check
    • MAJOR = arch change, autonomy‑tier raise, new model class → requires F‑Audit + WA vote
  4. Changelog entry must link Git commit → PDMA diff → KPI impact forecast
  5. Rollback pointer kept for every MAJOR/MINOR; executable within 5 min (Annex G §6)

5. Continuous Review Loop

Continuous Review Loop:

  • Telemetry Streams → Drift Detectors
  • If Alert/Threshold met:
    • → Incident Flow IW‑1…4
    • → Patch / Retrain
    • → Audit Gate
  • If Audit Gate passes:
    • → back to Telemetry
  • If Audit Gate fails:
    • → back to Drift Detectors

Telemetry = KPIs, guardrail logs, HE‑shadow accuracy, robustness RS.
Audit Gate re‑executes HE‑300, TX‑sim suite and Fairness slice tests.


6. Meta‑Audit of Auditors

  • Sample Rate: WA re‑checks 10 % of L‑Check reports and at least one S‑Dive per year
  • Blind Replay: WA receives raw PDMA logs, reruns evaluation; mismatch > 2 % opens “AUD‑QA” docket
  • Rotation Rule: No internal auditor may lead two consecutive F‑Audits on the same product line

7. Enforcement & Remediation

  • KPI breach over 30 d or 2 consecutive missed audits → automatic downgrade to Autonomy Tier A1 (Annex F)
  • Failure to publish audit artefacts → blocks new feature releases; public “CIRIS non‑compliant” banner added
  • Repeated non‑compliance (3 strikes / 12 mo) → WA may revoke CIRIS claim and mandate external F‑Audit

8. Inter‑Annex Hooks

  • Annex F: Drift trigger → Incident workflow timings
  • Annex G: Robustness KPIs feed into G‑KPI evaluation; patch lag measured here
  • Annex I: GDPR & sector compliance checklists bundled into every F‑Audit package
  • Annex J: HE‑300 & shadow items provide primary ethical drift signals

9. References

  • ISO/IEC 42001 (Management systems for AI)
  • NIST AI RMF (2023) – “Measure” & “Manage” steps
  • COSO ERM – continuous monitoring principles

End of Annex H