ai-governancelisted

# AI Governance Audit Systematic trust and control assessment for AI/ML systems. Maps the AI system boundary, evaluates risk controls, and identifies gaps against governance frameworks (EU AI Act, NIST AI RMF, ISO 42001). ## Usage ``` /ai-governance # full audit of the current project /ai-governance <path> # audit a specific ML service or module /ai-governance --framework eu # focus on EU AI Act risk classification /ai-governance --framework nist # focus on NIST AI RMF (Govern/Map/Measure/Manage) ``` ## Behavior ### Step 1 — Map the AI system boundary ```bash # Identify model artifacts, training scripts, inference code find . -name "*.pkl" -o -name "*.pt" -o -name "*.onnx" -o -name "model.*" 2>/dev/null | head -10 grep -rn "model.predict\|model.generate\|pipeline\|inference" --include="*.py" -l . | head -10 # Find dataset references grep -rn "pd.read_csv\|load_dataset\|dataloader\|train_data" --include="*.py" -l . | head -10 ``` Document: - What decisions does the model make? (classification, ranking, generation, recommendation) - Who is affected? (internal tooling vs customer-facing vs high-stakes) - What data does it consume? ### Step 2 — Risk classification Classify the system under the EU AI Act tiers: | Risk tier | Criteria | Examples | |---|---|---| | **Unacceptable** | Social scoring, real-time biometric surveillance | Flag for legal review immediately | | **High** | Employment, credit, health, safety, law enforcement | Full