borghei/Claude-Skills / ra-qm-team/eu-ai-act-specialist
eu-ai-act-specialist
VerifiedEU AI Act (Regulation EU 2024/1689) compliance specialist. Use for AI system risk classification, provider/deployer obligations, GPAI model compliance, conformity assessments, bias and fairness testing, and AI governance programs.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 28.5 KB |
| references/ai-act-classification-guide.md | markdown | 28.2 KB |
| references/ai-governance-framework.md | markdown | 32.1 KB |
| references/ai-technical-documentation-templates.md | markdown | 70.7 KB |
| scripts/ai_bias_detector.py | script | 31.9 KB |
| scripts/ai_compliance_checker.py | script | 55.3 KB |
| scripts/ai_risk_classifier.py | script | 42.3 KB |
references/ai-technical-documentation-templates.md
View on GitHub# AI Technical Documentation Templates
Production-ready templates for EU AI Act (Regulation EU 2024/1689) compliance documentation. Covers all required documentation for high-risk AI systems, GPAI models, and deployer obligations.
---
## Table of Contents
- [Template 1: AI System Description](#template-1-ai-system-description)
- [Template 2: Risk Management System Documentation](#template-2-risk-management-system-documentation)
- [Template 3: Data Governance Documentation](#template-3-data-governance-documentation)
- [Template 4: Testing and Validation Report](#template-4-testing-and-validation-report)
- [Template 5: Human Oversight Procedures](#template-5-human-oversight-procedures)
- [Template 6: Post-Market Monitoring Plan](#template-6-post-market-monitoring-plan)
- [Template 7: Incident Reporting Template](#template-7-incident-reporting-template)
- [Template 8: Fundamental Rights Impact Assessment](#template-8-fundamental-rights-impact-assessment)
---
## Template 1: AI System Description
```
═══════════════════════════════════════════════════════════════════
AI SYSTEM TECHNICAL DESCRIPTION
Regulation (EU) 2024/1689 — Art. 11, Annex IV
═══════════════════════════════════════════════════════════════════
Document ID: [TDS-XXXX-VX.X]
System Name: [Name]
System Version: [X.X.X]
Provider: [Legal entity name]
Date: [DD/MM/YYYY]
Classification: [HIGH-RISK / LIMITED / MINIMAL]
Annex III Category: [Category number and title, if applicable]
Document Status: [Draft / Under Review / Approved / Superseded]
Approved By: [Name, Title]
───────────────────────────────────────────────────────────────────
1. GENERAL DESCRIPTION
───────────────────────────────────────────────────────────────────
1.1 Provider Information
- Legal name:
- Registered address:
- Contact person:
- Contact email:
- Contact phone:
- EU Authorised Representative (if non-EU provider):
1.2 AI System Identification
- Trade name / brand:
- Model / type designation:
- Version / build:
- Hardware/software platform:
- EU Database registration number:
1.3 Intended Purpose
- Primary intended purpose:
- Specific use cases:
- Target deployer profile:
- Geographic scope of deployment:
- Deployment environment:
1.4 Persons Affected
- Categories of natural persons affected:
- Estimated number of affected persons:
- Vulnerable groups that may be affected:
- Nature of impact on affected persons:
1.5 Contraindications and Foreseeable Misuse
- Known contraindications (when NOT to use):
- Foreseeable misuse scenarios:
- Safeguards against misuse:
- Prohibited uses:
───────────────────────────────────────────────────────────────────
2. TECHNICAL ARCHITECTURE
───────────────────────────────────────────────────────────────────
2.1 System Architecture Overview
- Architecture diagram: [Reference to diagram]
- Component list:
Component 1: [Name, function, technology]
Component 2: [Name, function, technology]
- External dependencies:
- Integration points:
2.2 AI Model Specification
- Model type: [CNN, Transformer, XGBoost, ensemble, etc.]
- Model architecture: [Detailed architecture description]
- Number of parameters:
- Input modalities: [text, image, tabular, audio, video, etc.]
- Output type: [classification, regression, generation, etc.]
- Output format: [probability scores, labels, text, etc.]
- Inference latency: [typical, p95, p99]
- Computational requirements:
Training: [GPU type, count, time]
Inference: [CPU/GPU requirements, memory]
2.3 Key Design Choices
| Design Decision | Choice Made | Rationale | Alternatives Considered |
|-----------------|-------------|-----------|------------------------|
| | | | |
2.4 Software and Dependencies
- Programming language(s):
- ML framework:
- Key libraries and versions:
- Operating system:
- Containerization (if applicable):
───────────────────────────────────────────────────────────────────
3. TRAINING AND DATA
───────────────────────────────────────────────────────────────────
3.1 Training Data
- Data sources:
- Data volume: [number of samples, data size]
- Data collection period:
- Data collection method:
- Geographic origin of data:
- Languages:
3.2 Data Characteristics
- Feature descriptions:
| Feature | Type | Description | Range/Values |
|---------|------|-------------|-------------|
| | | | |
- Label / target variable:
- Label distribution:
3.3 Data Preparation
- Cleaning procedures:
- Annotation process:
- Labelling methodology:
- Quality assurance for labels:
- Data augmentation (if applied):
- Train/validation/test split ratios:
- Stratification method:
3.4 Data Governance (Art. 10 Summary)
- Bias examination results: [Reference to full report]
- Representativeness assessment:
- Special category data: [Yes/No; if yes, safeguards]
- Data governance documentation: [Reference to full document]
───────────────────────────────────────────────────────────────────
4. PERFORMANCE AND LIMITATIONS
───────────────────────────────────────────────────────────────────
4.1 Performance Metrics
| Metric | Overall | Group 1 | Group 2 | Group N |
|--------|---------|---------|---------|---------|
| | | | | |
4.2 Performance Conditions
- Optimal operating conditions:
- Degraded performance conditions:
- Failure conditions:
4.3 Known Limitations
- Technical limitations:
- Data limitations:
- Environmental limitations:
- Edge cases with degraded performance:
4.4 Benchmarking
- Benchmark datasets used:
- Comparison with state of the art:
- Comparison with human performance (if relevant):
───────────────────────────────────────────────────────────────────
5. RISK CLASSIFICATION
───────────────────────────────────────────────────────────────────
5.1 Classification Determination
- AI Act classification: [HIGH-RISK / LIMITED / MINIMAL]
- Annex III category (if applicable):
- Art. 6(3) exception analysis (if applicable):
- Classification rationale:
5.2 Applicable Obligations
| Obligation | Article | Compliance Status |
|-----------|---------|-------------------|
| | | |
5.3 Cross-References
- Risk Management System: [Document reference]
- Data Governance: [Document reference]
- Human Oversight Procedures: [Document reference]
- Post-Market Monitoring Plan: [Document reference]
───────────────────────────────────────────────────────────────────
6. CHANGE HISTORY
───────────────────────────────────────────────────────────────────
| Version | Date | Author | Changes | Impact Assessment |
|---------|------|--------|---------|-------------------|
| | | | | |
───────────────────────────────────────────────────────────────────
7. APPROVALS
───────────────────────────────────────────────────────────────────
| Role | Name | Signature | Date |
|------|------|-----------|------|
| Technical Lead | | | |
| AI Compliance Officer | | | |
| Quality Manager | | | |
```
---
## Template 2: Risk Management System Documentation
```
═══════════════════════════════════════════════════════════════════
RISK MANAGEMENT SYSTEM — AI SYSTEM
Regulation (EU) 2024/1689 — Art. 9
═══════════════════════════════════════════════════════════════════
Document ID: [RMS-XXXX-VX.X]
AI System: [System name and version]
Risk Management Lead: [Name, Title]
Date: [DD/MM/YYYY]
Review Date: [DD/MM/YYYY]
Document Status: [Draft / Under Review / Approved]
───────────────────────────────────────────────────────────────────
1. SCOPE AND OBJECTIVES
───────────────────────────────────────────────────────────────────
1.1 Scope
- AI system(s) covered:
- Lifecycle phases covered:
- Boundaries and exclusions:
1.2 Objectives
- Identify and manage risks to health, safety, and fundamental rights
- Ensure residual risks are acceptable
- Maintain risk management throughout system lifecycle
- Comply with Art. 9 requirements
1.3 Methodology
- Risk identification method: [FMEA, HAZOP, FTA, brainstorming, etc.]
- Risk evaluation criteria:
- Risk acceptability criteria:
- Review and update triggers:
───────────────────────────────────────────────────────────────────
2. RISK IDENTIFICATION
───────────────────────────────────────────────────────────────────
2.1 Intended Use Risks
| Risk ID | Hazard | Harm | Affected Persons | Source |
|---------|--------|------|------------------|--------|
| R-001 | | | | |
| R-002 | | | | |
2.2 Foreseeable Misuse Risks
| Risk ID | Misuse Scenario | Hazard | Harm | Affected Persons |
|---------|----------------|--------|------|------------------|
| R-M001 | | | | |
2.3 Fundamental Rights Risks
| Risk ID | Right Affected | Risk Description | Affected Groups |
|---------|---------------|------------------|-----------------|
| R-FR001 | Non-discrimination | | |
| R-FR002 | Privacy | | |
| R-FR003 | Freedom of expression | | |
───────────────────────────────────────────────────────────────────
3. RISK EVALUATION
───────────────────────────────────────────────────────────────────
3.1 Probability Scale
| Level | Probability | Description |
|-------|-------------|-------------|
| 1 | Rare | Less than 1 in 10,000 uses |
| 2 | Unlikely | 1 in 1,000 to 1 in 10,000 |
| 3 | Possible | 1 in 100 to 1 in 1,000 |
| 4 | Likely | 1 in 10 to 1 in 100 |
| 5 | Very likely | More than 1 in 10 |
3.2 Severity Scale
| Level | Severity | Description |
|-------|----------|-------------|
| 1 | Negligible | Minor inconvenience, easily reversible |
| 2 | Minor | Temporary harm, recoverable without intervention |
| 3 | Moderate | Significant harm, requires intervention to recover |
| 4 | Serious | Serious harm to health, safety, or rights |
| 5 | Critical | Death, permanent disability, irreversible rights violation |
3.3 Risk Matrix
| Probability \ Severity | Negligible | Minor | Moderate | Serious | Critical |
|----------------------|------------|-------|----------|---------|----------|
| Very likely | Medium | High | High | Unacceptable | Unacceptable |
| Likely | Low | Medium| High | High | Unacceptable |
| Possible | Low | Medium| Medium | High | High |
| Unlikely | Low | Low | Medium | Medium | High |
| Rare | Low | Low | Low | Medium | Medium |
3.4 Risk Evaluation Results
| Risk ID | Probability | Severity | Risk Level | Acceptable? |
|---------|-------------|----------|------------|-------------|
| R-001 | | | | |
───────────────────────────────────────────────────────────────────
4. RISK CONTROL MEASURES
───────────────────────────────────────────────────────────────────
| Risk ID | Control Measure | Type | Implementation | Verification | Status |
|---------|----------------|------|----------------|-------------|--------|
| R-001 | | [Elimination/Reduction/Protection] | | | |
Control type hierarchy (most to least preferred):
1. Elimination — remove the source of risk
2. Reduction — reduce probability or severity
3. Protection — safeguards, warnings, training
───────────────────────────────────────────────────────────────────
5. RESIDUAL RISK ASSESSMENT
───────────────────────────────────────────────────────────────────
5.1 Individual Residual Risks
| Risk ID | Residual Probability | Residual Severity | Residual Level | Acceptable? |
|---------|---------------------|-------------------|----------------|-------------|
| R-001 | | | | |
5.2 Overall Residual Risk
- Overall residual risk determination: [Acceptable / Not acceptable]
- Risk-benefit analysis:
Benefits: [Description of system benefits]
Residual risks: [Summary of remaining risks]
Determination: [Benefits outweigh residual risks / Do not outweigh]
- Comparison with state of the art:
5.3 Information for Safety
- Residual risks communicated in instructions for use:
- Training requirements for deployers:
- Limitations highlighted:
───────────────────────────────────────────────────────────────────
6. TESTING
───────────────────────────────────────────────────────────────────
6.1 Pre-Market Testing
| Test ID | Test Description | Metrics | Threshold | Result | Pass/Fail |
|---------|-----------------|---------|-----------|--------|-----------|
| | | | | | |
6.2 Real-World Conditions Testing (if applicable)
- Test environment description:
- Test duration:
- Test population:
- Results summary:
───────────────────────────────────────────────────────────────────
7. REVIEW AND UPDATE PLAN
───────────────────────────────────────────────────────────────────
- Scheduled review frequency: [Annually / Semi-annually / Quarterly]
- Update triggers:
[ ] Serious incident reported
[ ] Post-market monitoring data indicates new risk
[ ] Significant system change (retraining, new data, new context)
[ ] New regulatory guidance or harmonised standard
[ ] Deployer feedback indicating issues
- Review record:
| Review Date | Reviewer | Changes Made | Rationale |
|-------------|----------|-------------|-----------|
| | | | |
───────────────────────────────────────────────────────────────────
8. APPROVALS
───────────────────────────────────────────────────────────────────
| Role | Name | Signature | Date |
|------|------|-----------|------|
| Risk Management Lead | | | |
| AI Compliance Officer | | | |
| Technical Lead | | | |
```
---
## Template 3: Data Governance Documentation
```
═══════════════════════════════════════════════════════════════════
DATA GOVERNANCE DOCUMENTATION — AI SYSTEM
Regulation (EU) 2024/1689 — Art. 10
═══════════════════════════════════════════════════════════════════
Document ID: [DGD-XXXX-VX.X]
AI System: [System name and version]
Data Governance Lead: [Name, Title]
Date: [DD/MM/YYYY]
Document Status: [Draft / Under Review / Approved]
───────────────────────────────────────────────────────────────────
1. DATA OVERVIEW
───────────────────────────────────────────────────────────────────
1.1 Dataset Inventory
| Dataset | Purpose | Source | Volume | Period | Format |
|---------|---------|--------|--------|--------|--------|
| Training | | | | | |
| Validation | | | | | |
| Testing | | | | | |
1.2 Data Types
- Personal data processed: [Yes/No]
- Special category data (Art. 9 GDPR): [Yes/No]
- If special category: justification for processing:
- Safeguards for special category data:
───────────────────────────────────────────────────────────────────
2. DESIGN CHOICES (Art. 10(2)(a))
───────────────────────────────────────────────────────────────────
2.1 Data Collection Design
- Why these data sources were chosen:
- Data sources rejected and reasons:
- Sampling methodology:
- Collection timeframe rationale:
- Geographic scope rationale:
2.2 Labelling Strategy
- Labelling methodology:
- Number of annotators:
- Inter-annotator agreement:
- Quality assurance process:
- Labelling guidelines: [Reference]
───────────────────────────────────────────────────────────────────
3. DATA PREPARATION (Art. 10(2)(b))
───────────────────────────────────────────────────────────────────
3.1 Processing Steps
| Step | Process | Tool/Method | Parameters | Quality Check |
|------|---------|-------------|-----------|---------------|
| 1 | Cleaning | | | |
| 2 | Filtering | | | |
| 3 | Annotation | | | |
| 4 | Enrichment | | | |
| 5 | Aggregation | | | |
3.2 Data Quality Metrics
| Metric | Value | Threshold | Pass/Fail |
|--------|-------|-----------|-----------|
| Completeness | | | |
| Accuracy | | | |
| Consistency | | | |
| Timeliness | | | |
| Duplication rate | | | |
───────────────────────────────────────────────────────────────────
4. ASSUMPTIONS (Art. 10(2)(c))
───────────────────────────────────────────────────────────────────
- What the data is meant to measure:
- What the data is meant to represent:
- Assumptions about data-to-reality mapping:
- Known gaps between data and real-world distribution:
- Limitations of the data:
───────────────────────────────────────────────────────────────────
5. SUFFICIENCY AND REPRESENTATIVENESS (Art. 10(2)(e)(f))
───────────────────────────────────────────────────────────────────
5.1 Data Sufficiency
- Total samples: [number]
- Sufficiency assessment methodology:
- Learning curve analysis (if performed):
- Conclusion on sufficiency:
5.2 Representativeness Analysis
| Attribute | Dataset Distribution | Population Distribution | Ratio | Assessment |
|-----------|---------------------|------------------------|-------|------------|
| | | | | |
- Population reference source:
- Overall representativeness assessment:
- Gaps identified:
- Mitigation for gaps:
───────────────────────────────────────────────────────────────────
6. BIAS EXAMINATION (Art. 10(2)(f))
───────────────────────────────────────────────────────────────────
6.1 Bias Analysis Methodology
- Tools used: [e.g., ai_bias_detector.py]
- Protected attributes examined:
- Metrics calculated:
- Thresholds applied:
6.2 Bias Analysis Results
| Attribute | Metric | Value | Threshold | Pass/Fail | Severity |
|-----------|--------|-------|-----------|-----------|----------|
| | | | | | |
6.3 Proxy Variable Analysis
| Feature | Protected Attribute | Correlation | Risk Level |
|---------|-------------------|-------------|------------|
| | | | |
6.4 Bias Examination Conclusion
- Summary of findings:
- Overall bias risk assessment:
- Reference to full bias detection report:
───────────────────────────────────────────────────────────────────
7. BIAS MITIGATION (Art. 10(2)(g))
───────────────────────────────────────────────────────────────────
| Bias Finding | Mitigation Measure | Type | Effectiveness | Residual Bias |
|-------------|-------------------|------|---------------|---------------|
| | | [Pre/In/Post-processing] | | |
- Mitigation verification results:
- Residual bias assessment:
- Residual bias acceptance rationale:
───────────────────────────────────────────────────────────────────
8. SPECIAL CATEGORY DATA (Art. 10(5))
───────────────────────────────────────────────────────────────────
(Complete only if special category data is processed)
- Special categories processed: [race, political opinions, religious
beliefs, trade union membership, genetic data, biometric data, health
data, sex life, sexual orientation]
- Strict necessity justification:
- Specific purpose: [bias detection and correction]
- Safeguards implemented:
[ ] Encryption at rest and in transit
[ ] Access controls (list authorized persons)
[ ] Purpose limitation enforcement
[ ] Data minimization applied
[ ] Retention limitation defined
[ ] DPIA completed
- Legal basis:
───────────────────────────────────────────────────────────────────
9. DATA MANAGEMENT
───────────────────────────────────────────────────────────────────
- Storage location:
- Access controls:
- Retention period:
- Deletion procedure:
- Version control:
- Data lineage tracking:
───────────────────────────────────────────────────────────────────
10. APPROVALS
───────────────────────────────────────────────────────────────────
| Role | Name | Signature | Date |
|------|------|-----------|------|
| Data Governance Lead | | | |
| Data Protection Officer | | | |
| AI Compliance Officer | | | |
```
---
## Template 4: Testing and Validation Report
```
═══════════════════════════════════════════════════════════════════
TESTING AND VALIDATION REPORT — AI SYSTEM
Regulation (EU) 2024/1689 — Art. 9(7), Art. 15
═══════════════════════════════════════════════════════════════════
Document ID: [TVR-XXXX-VX.X]
AI System: [System name and version]
Test Lead: [Name, Title]
Test Period: [Start date — End date]
Document Status: [Draft / Under Review / Approved]
───────────────────────────────────────────────────────────────────
1. TEST PLAN SUMMARY
───────────────────────────────────────────────────────────────────
1.1 Objectives
- Verify system meets accuracy requirements (Art. 15)
- Verify robustness against errors and adversarial inputs (Art. 15)
- Verify cybersecurity measures (Art. 15)
- Validate risk management measures (Art. 9)
- Verify fairness across demographic groups (Art. 10)
1.2 Test Environment
- Hardware:
- Software:
- Configuration:
- Differences from production environment:
1.3 Test Data
- Test dataset description:
- Test dataset size:
- Test data source:
- Relationship to training data: [Independent / Holdout / Cross-validated]
───────────────────────────────────────────────────────────────────
2. ACCURACY TESTING (Art. 15(1))
───────────────────────────────────────────────────────────────────
2.1 Overall Performance
| Metric | Value | Threshold | Pass/Fail |
|--------|-------|-----------|-----------|
| Accuracy | | | |
| Precision | | | |
| Recall | | | |
| F1 Score | | | |
| AUC-ROC | | | |
| [Other] | | | |
2.2 Performance by Demographic Group
| Group | Metric 1 | Metric 2 | Metric 3 | Threshold | Pass/Fail |
|-------|----------|----------|----------|-----------|-----------|
| | | | | | |
2.3 Performance by Use Case / Scenario
| Scenario | Metric | Value | Threshold | Pass/Fail |
|----------|--------|-------|-----------|-----------|
| | | | | |
2.4 Accuracy Conclusions
- Overall accuracy assessment:
- Performance variations across groups:
- Accuracy level to declare in instructions for use:
───────────────────────────────────────────────────────────────────
3. ROBUSTNESS TESTING (Art. 15(2)(3))
───────────────────────────────────────────────────────────────────
3.1 Input Error Resilience
| Error Type | Test Method | Degradation | Acceptable? |
|-----------|-------------|-------------|-------------|
| Missing values | | | |
| Noisy input | | | |
| Out-of-range values | | | |
| Corrupted input | | | |
| Wrong format | | | |
3.2 Edge Case Testing
| Edge Case | Input Description | Expected Behaviour | Actual Behaviour | Pass/Fail |
|-----------|------------------|-------------------|-----------------|-----------|
| | | | | |
3.3 Fail-Safe Behaviour
| Failure Scenario | Expected Fail-Safe | Actual Behaviour | Pass/Fail |
|-----------------|-------------------|-----------------|-----------|
| | | | |
3.4 Redundancy Verification
| Component | Redundancy Measure | Test Method | Result |
|-----------|-------------------|-------------|--------|
| | | | |
───────────────────────────────────────────────────────────────────
4. CYBERSECURITY TESTING (Art. 15(4)(5))
───────────────────────────────────────────────────────────────────
4.1 Data Poisoning Resistance
| Attack Type | Test Method | Result | Mitigation |
|------------|-------------|--------|------------|
| Label flipping | | | |
| Backdoor insertion | | | |
| Data injection | | | |
4.2 Adversarial Example Resistance
| Attack Method | Perturbation Level | Success Rate | Mitigation |
|--------------|-------------------|-------------|------------|
| FGSM | | | |
| PGD | | | |
| [Other] | | | |
4.3 Model Integrity
| Test | Method | Result | Pass/Fail |
|------|--------|--------|-----------|
| Model tampering detection | | | |
| Model extraction resistance | | | |
| Membership inference resistance | | | |
4.4 Infrastructure Security
| Control | Test Method | Result | Pass/Fail |
|---------|-------------|--------|-----------|
| API authentication | | | |
| Data encryption | | | |
| Access logging | | | |
| Network segmentation | | | |
───────────────────────────────────────────────────────────────────
5. FAIRNESS TESTING (Art. 10)
───────────────────────────────────────────────────────────────────
5.1 Fairness Metrics
| Attribute | Demographic Parity | Equalized Odds | Predictive Parity | Pass/Fail |
|-----------|-------------------|----------------|-------------------|-----------|
| | | | | |
5.2 Disparate Impact Analysis
| Attribute | Advantaged Group | Disadvantaged Group | Impact Ratio | Pass/Fail |
|-----------|-----------------|---------------------|-------------|-----------|
| | | | | |
5.3 Fairness Conclusions
- Overall fairness assessment:
- Residual disparities:
- Mitigation measures applied:
───────────────────────────────────────────────────────────────────
6. OVERALL TEST CONCLUSIONS
───────────────────────────────────────────────────────────────────
- Total tests executed:
- Tests passed:
- Tests failed:
- Overall determination: [PASS / CONDITIONAL PASS / FAIL]
- Conditions (if conditional pass):
- Recommended actions:
───────────────────────────────────────────────────────────────────
7. APPROVALS
───────────────────────────────────────────────────────────────────
| Role | Name | Signature | Date |
|------|------|-----------|------|
| Test Lead | | | |
| AI Compliance Officer | | | |
| Quality Manager | | | |
```
---
## Template 5: Human Oversight Procedures
```
═══════════════════════════════════════════════════════════════════
HUMAN OVERSIGHT PROCEDURES — AI SYSTEM
Regulation (EU) 2024/1689 — Art. 14
═══════════════════════════════════════════════════════════════════
Document ID: [HOP-XXXX-VX.X]
AI System: [System name and version]
Oversight Level: [Human-in-the-loop / Human-on-the-loop / Human-in-command]
Date: [DD/MM/YYYY]
Document Status: [Draft / Under Review / Approved]
───────────────────────────────────────────────────────────────────
1. OVERSIGHT DESIGN
───────────────────────────────────────────────────────────────────
1.1 Oversight Level Determination
- Selected level: [Human-in-the-loop / Human-on-the-loop / Human-in-command]
- Rationale for selection:
- Risk factors considered:
- Decision impact assessment:
1.2 Oversight Architecture
- Where in the decision pipeline oversight occurs:
- What decisions require human approval:
- What decisions can proceed automatically:
- Confidence threshold for automatic processing (if applicable):
- Escalation criteria:
───────────────────────────────────────────────────────────────────
2. OVERSIGHT PERSONNEL
───────────────────────────────────────────────────────────────────
2.1 Roles
| Role | Responsibility | Authority Level | Required Number |
|------|---------------|-----------------|-----------------|
| Primary Overseer | | | |
| Backup Overseer | | | |
| Escalation Authority | | | |
2.2 Required Competencies
| Competency | Description | Assessment Method |
|-----------|-------------|-------------------|
| Domain knowledge | | |
| AI system understanding | | |
| Output interpretation | | |
| Bias awareness | | |
| Decision authority | | |
2.3 Training Requirements
| Training Module | Duration | Frequency | Certification |
|----------------|----------|-----------|---------------|
| System operation | | | |
| Output interpretation | | | |
| Override procedures | | | |
| Bias awareness | | | |
| Incident reporting | | | |
2.4 Training Records
- Training records maintained by:
- Training effectiveness assessment method:
- Recertification period:
───────────────────────────────────────────────────────────────────
3. MONITORING PROCEDURES
───────────────────────────────────────────────────────────────────
3.1 Monitoring Interface
- Dashboard description:
- Key indicators displayed:
- Alert mechanism:
- Real-time vs batch monitoring:
3.2 Key Performance Indicators
| KPI | Description | Threshold | Alert Level |
|-----|------------|-----------|------------|
| | | | |
3.3 Review Procedures
- Review frequency: [Real-time / Hourly / Daily / Weekly]
- Sample review rate: [percentage of decisions reviewed]
- Review documentation requirements:
- Quality assurance for reviews:
───────────────────────────────────────────────────────────────────
4. INTERVENTION PROCEDURES
───────────────────────────────────────────────────────────────────
4.1 Override Mechanism
- How to override a system decision:
- Override authority levels:
- Override documentation requirements:
- Override logging:
4.2 Stop / Shutdown Procedure
- Emergency stop mechanism: [Button location / Command / API]
- Graceful shutdown procedure:
- Emergency shutdown procedure:
- System state after shutdown:
- Restart procedure:
4.3 Decision Reversal Process
- How to reverse a system decision:
- Time window for reversal:
- Notification to affected persons:
- Documentation requirements:
4.4 Escalation Path
| Trigger | Escalation To | Response Time | Action |
|---------|--------------|---------------|--------|
| | | | |
───────────────────────────────────────────────────────────────────
5. AUTOMATION BIAS SAFEGUARDS (Art. 14(4)(b))
───────────────────────────────────────────────────────────────────
5.1 Safeguards Implemented
| Safeguard | Description | Implementation |
|-----------|-------------|----------------|
| Mandatory review | Overseers must actively confirm decisions | |
| Randomized verification | Random subset requires independent verification | |
| Confidence display | System confidence shown alongside decisions | |
| Dissenting information | Counter-evidence actively presented | |
| Rotation | Overseers rotated to prevent complacency | |
| Calibration exercises | Regular exercises comparing AI vs correct answers | |
5.2 Effectiveness Measurement
- How automation bias is measured:
- Override rate monitoring:
- Correlation between AI confidence and human agreement:
- Periodic assessment schedule:
───────────────────────────────────────────────────────────────────
6. RECORD-KEEPING
───────────────────────────────────────────────────────────────────
- Oversight actions logged: [Yes]
- Override decisions logged with rationale: [Yes]
- Review outcomes documented: [Yes]
- Incident reports linked: [Yes]
- Retention period:
- Storage location:
───────────────────────────────────────────────────────────────────
7. APPROVALS
───────────────────────────────────────────────────────────────────
| Role | Name | Signature | Date |
|------|------|-----------|------|
| Oversight Coordinator | | | |
| AI Compliance Officer | | | |
```
---
## Template 6: Post-Market Monitoring Plan
```
═══════════════════════════════════════════════════════════════════
POST-MARKET MONITORING PLAN — AI SYSTEM
Regulation (EU) 2024/1689 — Art. 72
═══════════════════════════════════════════════════════════════════
Document ID: [PMP-XXXX-VX.X]
AI System: [System name and version]
Monitoring Lead: [Name, Title]
Date: [DD/MM/YYYY]
Document Status: [Draft / Under Review / Approved]
───────────────────────────────────────────────────────────────────
1. MONITORING SCOPE
───────────────────────────────────────────────────────────────────
- AI system(s) covered:
- Monitoring start date:
- Geographic scope:
- Number of deployers:
- Estimated number of affected persons:
───────────────────────────────────────────────────────────────────
2. DATA COLLECTION
───────────────────────────────────────────────────────────────────
2.1 Data Sources
| Source | Data Type | Collection Method | Frequency |
|--------|-----------|------------------|-----------|
| System logs | Performance metrics | Automated | Continuous |
| Deployer feedback | Incident reports, complaints | Manual | As received |
| User feedback | Satisfaction, accuracy | Survey | Quarterly |
| Market surveillance | Authority communications | Manual | As received |
| Scientific literature | New risks, state of art | Manual review | Semi-annual |
| Media monitoring | Public reports, incidents | Automated | Continuous |
2.2 Performance Metrics Monitored
| Metric | Baseline | Alert Threshold | Critical Threshold | Frequency |
|--------|----------|----------------|-------------------|-----------|
| Accuracy | | | | |
| Fairness | | | | |
| Drift | | | | |
| Latency | | | | |
| Error rate | | | | |
───────────────────────────────────────────────────────────────────
3. ANALYSIS PROCEDURES
───────────────────────────────────────────────────────────────────
- Performance trend analysis: [Method, frequency]
- Bias monitoring analysis: [Method, frequency]
- Drift detection: [Method, frequency]
- Root cause analysis for anomalies: [Method]
- Cross-system interaction analysis: [Method]
───────────────────────────────────────────────────────────────────
4. RISK MANAGEMENT INTEGRATION
───────────────────────────────────────────────────────────────────
- Trigger for risk management update:
- Process for updating risk assessment with new data:
- Feedback loop to development team:
- Change management for system modifications:
───────────────────────────────────────────────────────────────────
5. INCIDENT REPORTING (Art. 73)
───────────────────────────────────────────────────────────────────
- Serious incident definition: [Reference Art. 3(49)]
- Reporting deadline: 15 days from awareness
- Reporting authority: [Market surveillance authority]
- Report format: [Reference to Incident Reporting Template]
- Responsible person:
- Escalation procedure:
───────────────────────────────────────────────────────────────────
6. CORRECTIVE ACTIONS
───────────────────────────────────────────────────────────────────
| Trigger | Corrective Action | Timeline | Authority |
|---------|------------------|----------|-----------|
| Performance below threshold | Investigation + remediation | 30 days | Monitoring Lead |
| Bias exceeding threshold | Bias mitigation + revalidation | 60 days | Data Governance Lead |
| Serious incident | Immediate corrective action | Immediate | AI Compliance Officer |
| Regulatory non-compliance | Compliance remediation | Per authority | AI Compliance Officer |
───────────────────────────────────────────────────────────────────
7. REPORTING
───────────────────────────────────────────────────────────────────
| Report | Audience | Frequency | Content |
|--------|----------|-----------|---------|
| Monitoring dashboard | Operations team | Real-time | KPIs, alerts |
| Monthly monitoring report | AI Compliance | Monthly | Trends, incidents, actions |
| Management review | Leadership | Quarterly | Summary, risks, recommendations |
| Annual post-market report | Authority (if required) | Annual | Comprehensive review |
───────────────────────────────────────────────────────────────────
8. APPROVALS
───────────────────────────────────────────────────────────────────
| Role | Name | Signature | Date |
|------|------|-----------|------|
| Monitoring Lead | | | |
| AI Compliance Officer | | | |
| Quality Manager | | | |
```
---
## Template 7: Incident Reporting Template
```
═══════════════════════════════════════════════════════════════════
AI SYSTEM INCIDENT REPORT
Regulation (EU) 2024/1689 — Art. 73
═══════════════════════════════════════════════════════════════════
REPORTING DEADLINE: 15 days from becoming aware of serious incident
───────────────────────────────────────────────────────────────────
1. INCIDENT IDENTIFICATION
───────────────────────────────────────────────────────────────────
Incident ID: [IR-XXXX]
Report Date: [DD/MM/YYYY]
Date of Awareness: [DD/MM/YYYY]
Date of Occurrence: [DD/MM/YYYY]
Reporting Deadline: [DD/MM/YYYY — 15 days from awareness]
Report Type: [ ] Initial [ ] Follow-up [ ] Final
Report Status: [Submitted / Acknowledged / Under review]
───────────────────────────────────────────────────────────────────
2. REPORTER INFORMATION
───────────────────────────────────────────────────────────────────
Reporter Role: [ ] Provider [ ] Deployer [ ] Authorised Rep
Organisation: [Legal name]
Contact Person: [Name]
Contact Email: [Email]
Contact Phone: [Phone]
Address: [Address]
───────────────────────────────────────────────────────────────────
3. AI SYSTEM IDENTIFICATION
───────────────────────────────────────────────────────────────────
System Name: [Name]
System Version: [Version]
Provider: [Provider name]
EU Database ID: [Registration number]
Risk Classification: [HIGH-RISK]
Annex III Category: [Category number]
CE Marking: [ ] Yes [ ] No
Notified Body: [Name and number, if applicable]
───────────────────────────────────────────────────────────────────
4. INCIDENT DESCRIPTION
───────────────────────────────────────────────────────────────────
4.1 Summary
[Clear, concise description of what happened]
4.2 Serious Incident Classification
The incident involves: (select all that apply)
[ ] Death of a natural person
[ ] Serious damage to health of a natural person
[ ] Serious and irreversible disruption of critical infrastructure
[ ] Breach of fundamental rights obligations
[ ] Serious damage to property
[ ] Serious damage to the environment
4.3 Detailed Description
- Sequence of events:
- AI system behaviour during incident:
- System inputs at time of incident:
- System outputs at time of incident:
- Human oversight actions taken:
- Environmental conditions:
4.4 Root Cause (known or suspected)
[ ] Model error / incorrect prediction
[ ] Data quality issue
[ ] System malfunction / software bug
[ ] Adversarial attack / manipulation
[ ] Misuse by deployer / user
[ ] Environmental factor
[ ] Unknown
Description:
───────────────────────────────────────────────────────────────────
5. IMPACT ASSESSMENT
───────────────────────────────────────────────────────────────────
5.1 Persons Affected
- Number of persons affected:
- Categories of affected persons:
- Vulnerable groups affected:
- Severity of harm:
5.2 Geographic Scope
- Member states where incident occurred:
- Member states where system is deployed:
5.3 Ongoing Risk
- Is the risk ongoing? [ ] Yes [ ] No
- Likelihood of recurrence:
- Number of similar systems potentially affected:
───────────────────────────────────────────────────────────────────
6. IMMEDIATE CORRECTIVE ACTIONS
───────────────────────────────────────────────────────────────────
| Action | Date Implemented | Status | Effectiveness |
|--------|-----------------|--------|---------------|
| | | | |
System status: [ ] Operating normally [ ] Operating with restrictions
[ ] Suspended [ ] Withdrawn from market
───────────────────────────────────────────────────────────────────
7. PLANNED CORRECTIVE ACTIONS
───────────────────────────────────────────────────────────────────
| Action | Target Date | Responsible | Status |
|--------|------------|-------------|--------|
| | | | |
───────────────────────────────────────────────────────────────────
8. NOTIFICATIONS
───────────────────────────────────────────────────────────────────
| Authority / Body | Date Notified | Reference Number |
|-----------------|---------------|-----------------|
| Market surveillance authority ([Country]) | | |
| Notified body (if applicable) | | |
| Data protection authority (if personal data) | | |
| Deployers | | |
───────────────────────────────────────────────────────────────────
9. ATTACHMENTS
───────────────────────────────────────────────────────────────────
[ ] System logs from incident period
[ ] Error reports
[ ] Photographs / screenshots
[ ] Witness statements
[ ] Medical reports (if applicable)
[ ] Technical analysis report
[ ] Other: ____________
───────────────────────────────────────────────────────────────────
10. DECLARATION
───────────────────────────────────────────────────────────────────
I confirm that the information provided is accurate and complete
to the best of my knowledge.
Name:
Title:
Signature:
Date:
```
---
## Template 8: Fundamental Rights Impact Assessment
```
═══════════════════════════════════════════════════════════════════
FUNDAMENTAL RIGHTS IMPACT ASSESSMENT (FRIA)
Regulation (EU) 2024/1689 — Art. 27
═══════════════════════════════════════════════════════════════════
Required for: Deployers that are public bodies or private entities
providing public services, using high-risk AI systems
Document ID: [FRIA-XXXX-VX.X]
AI System: [System name and version]
Deployer: [Organisation name]
Assessment Lead: [Name, Title]
Date: [DD/MM/YYYY]
Document Status: [Draft / Under Review / Approved / Submitted]
───────────────────────────────────────────────────────────────────
1. DEPLOYER PROCESSES (Art. 27(1)(a))
───────────────────────────────────────────────────────────────────
1.1 Description of Use
- Business process where AI system is used:
- Role of AI system in the process:
- Decision types supported by the system:
- Human involvement in decision process:
- Volume of decisions processed:
1.2 Operational Context
- Department / unit using the system:
- Geographic deployment:
- Regulatory context:
- Public service nature of the use:
───────────────────────────────────────────────────────────────────
2. USAGE PARAMETERS (Art. 27(1)(b))
───────────────────────────────────────────────────────────────────
- Period of intended use: [Start date — End date / Ongoing]
- Frequency of use: [Continuous / Daily / Weekly / As needed]
- Volume: [Number of decisions per period]
- Geographic scope: [Cities / Regions / National]
───────────────────────────────────────────────────────────────────
3. AFFECTED PERSONS AND GROUPS (Art. 27(1)(c))
───────────────────────────────────────────────────────────────────
3.1 Categories of Affected Persons
| Category | Estimated Number | Vulnerability Factors | Exposure Level |
|----------|-----------------|----------------------|----------------|
| | | | [High/Med/Low] |
3.2 Specific Groups Likely to Be Affected
| Group | Why Affected | Potential Disproportionate Impact | Mitigation |
|-------|-------------|----------------------------------|------------|
| Children / minors | | | |
| Elderly persons | | | |
| Persons with disabilities | | | |
| Ethnic minorities | | | |
| Low-income persons | | | |
| Migrants / refugees | | | |
| [Other groups] | | | |
───────────────────────────────────────────────────────────────────
4. SPECIFIC RISKS OF HARM (Art. 27(1)(d))
───────────────────────────────────────────────────────────────────
4.1 Fundamental Rights Risk Assessment
| Right | Risk Description | Affected Groups | Likelihood | Severity | Risk Level |
|-------|-----------------|-----------------|------------|----------|------------|
| **Human dignity** (Art. 1 CFR) | | | | | |
| **Right to life** (Art. 2 CFR) | | | | | |
| **Integrity of the person** (Art. 3 CFR) | | | | | |
| **Respect for private life** (Art. 7 CFR) | | | | | |
| **Protection of personal data** (Art. 8 CFR) | | | | | |
| **Non-discrimination** (Art. 21 CFR) | | | | | |
| **Equality between women and men** (Art. 23 CFR) | | | | | |
| **Rights of the child** (Art. 24 CFR) | | | | | |
| **Integration of persons with disabilities** (Art. 26 CFR) | | | | | |
| **Right to an effective remedy** (Art. 47 CFR) | | | | | |
| **Right to good administration** (Art. 41 CFR) | | | | | |
| **Freedom of expression** (Art. 11 CFR) | | | | | |
| **Freedom of assembly** (Art. 12 CFR) | | | | | |
| **Right to education** (Art. 14 CFR) | | | | | |
| **Workers' rights** (Art. 27-32 CFR) | | | | | |
| **Consumer protection** (Art. 38 CFR) | | | | | |
CFR = Charter of Fundamental Rights of the European Union
4.2 Cumulative and Systemic Risks
- Risk of cumulative impact across multiple uses:
- Risk of systemic discrimination:
- Risk of chilling effects on rights exercise:
───────────────────────────────────────────────────────────────────
5. HUMAN OVERSIGHT MEASURES (Art. 27(1)(e))
───────────────────────────────────────────────────────────────────
- Oversight level implemented:
- Oversight personnel qualifications:
- Override procedures:
- Escalation procedures:
- Automation bias safeguards:
- Assessment of oversight adequacy:
───────────────────────────────────────────────────────────────────
6. RISK MITIGATION MEASURES (Art. 27(1)(f))
───────────────────────────────────────────────────────────────────
| Risk Identified | Mitigation Measure | Implementation Status | Effectiveness |
|----------------|-------------------|----------------------|---------------|
| | | | |
- Additional safeguards for vulnerable groups:
- Accessibility measures:
- Information provided to affected persons:
- Alternative (non-AI) pathway available: [ ] Yes [ ] No
If yes, describe:
───────────────────────────────────────────────────────────────────
7. GOVERNANCE AND COMPLAINTS (Art. 27(1)(g))
───────────────────────────────────────────────────────────────────
7.1 Governance Mechanisms
- Internal governance structure for AI use:
- Decision-making authority for AI deployment:
- Regular review schedule:
- Accountability framework:
7.2 Complaint Mechanisms
- How affected persons can raise concerns:
- Complaint handling process:
- Timeline for complaint resolution:
- Appeal mechanism:
- Contact information for complaints:
7.3 Redress
- How affected persons can seek redress:
- Types of redress available:
- Information provided to affected persons about rights:
───────────────────────────────────────────────────────────────────
8. CONCLUSIONS AND DETERMINATION
───────────────────────────────────────────────────────────────────
- Overall fundamental rights risk level:
[ ] Low — risks adequately mitigated
[ ] Medium — residual risks exist with mitigation in place
[ ] High — significant residual risks; additional measures needed
[ ] Unacceptable — deployment should not proceed
- Determination: [ ] Proceed with deployment
[ ] Proceed with conditions
[ ] Do not proceed
- Conditions (if applicable):
- Review and update schedule:
───────────────────────────────────────────────────────────────────
9. MARKET SURVEILLANCE AUTHORITY SUBMISSION
───────────────────────────────────────────────────────────────────
- Authority name:
- Submission date:
- Submission method:
- Reference number:
- Authority response (if received):
───────────────────────────────────────────────────────────────────
10. APPROVALS
───────────────────────────────────────────────────────────────────
| Role | Name | Signature | Date |
|------|------|-----------|------|
| Assessment Lead | | | |
| Legal Counsel | | | |
| Data Protection Officer | | | |
| Senior Leadership | | | |
```
---
**Regulation Reference:** Regulation (EU) 2024/1689 — EU Artificial Intelligence Act
**Last Updated:** March 2026
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |