⚙️ Model Governance SOP

APJ Enterprise LLC — NIST AI RMF v1.0 | ISO/IEC 42001 (in progress)

1. Registration & Documentation

• Register each model with: purpose, data sources, training date, hyperparameters, owners, version, and license.
• Maintain a Model Card (description, metrics, limitations).

2. Bias & Fairness Testing

• Use representative test sets; evaluate disparate impact where applicable.
• Record results and mitigations.

3. Change Control

• Retraining requires ticket, approval, and rollback plan.
• Version every model artifact; retain training data hashes.

4. Model Review Checklist

| Item | Reviewer | Status | Notes | |—|—|—|—| | Model Card completed | | | | | Dataset consent & rights verified | | | | | Bias tests executed & passed thresholds | | | | | Security review (secrets, injection) | | | | | Monitoring metrics configured | | | |

---

Owner: Data Science Lead • Last Updated: Tuesday, October 7, 2025