| name | eu-ai-act-compliance |
| description | EU AI Act compliance logging and requirements. Use this skill when ensuring transparency, human oversight, and record-keeping per Regulation (EU) 2024/1689 — even if they just say "add compliance logging" or "make sure this is EU AI Act compliant". Not for security-code-auditor, privacy-first. |
| version | 0.2.10 |
| category | compliance |
| metadata | {"author":"<ORG_NAME>","spec":"agentskills.io","regulation":"Regulation (EU) 2024/1689","effective_date":"2026-08-02T00:00:00.000Z"} |
| license | MIT |
EU AI Act Compliance
When to Use
- User asks to ensure EU AI Act compliance for logging or transparency
- Need to implement human oversight or record-keeping per Regulation (EU) 2024/1689
- Even if they just say "add compliance logging" or "make sure this is EU AI Act compliant"
Quick Start
import { AIActLogger } from "./eu-ai-act-compliance";
const logger = new AIActLogger({
systemId: "<PROJECT_ID>",
providerName: "<ORG_NAME>",
riskClassification: "limited_risk",
});
await logger.logOperation({
operation: "ai_inference",
inputData: { source: "user_request", hash: "sha256:abc123..." },
outputData: { result: "example_result", confidence: 0.85 },
humanOversight: { reviewerId: "user_123", decision: "approved", timestamp: new Date().toISOString() },
});
Core Concepts
| Concept | Article | Description |
|---|
| Automatic Logging | Art. 12 | Record events over system lifetime |
| Transparency | Art. 50 | Disclose AI interaction to users |
| Human Oversight | Art. 14 | Enable human intervention |
| Data Governance | Art. 10 | Document training/validation data |
| Retention | Art. 19 | Keep logs minimum 6 months |
Risk Classification
Limited Risk (Article 50)
Systems interacting with natural persons must:
- Disclose AI interaction at first contact (Art. 50.1).
- Mark synthetic content as AI-generated (Art. 50.2).
High Risk (Chapter III, Articles 8-17)
Systems in Annex III (recruitment, credit scoring, etc.) require:
- Risk management (Art. 9) & Data governance (Art. 10).
- Technical documentation (Art. 11) & Automatic logging (Art. 12).
- Human oversight design (Art. 14) & CE marking (Art. 48).
Logging Requirements (Article 12)
interface AIActLogEntry {
timestamp: string;
systemId: string;
operation: string;
inputData: { source: string; hash: string; description: string };
outputData: { result: string; confidence?: number; explanation?: string };
humanOversight?: { reviewerId: string; decision: "approved" | "rejected" | "modified"; timestamp: string };
retentionUntil: string;
}
Compliance Checklist
References
references/eu-ai-act-articles.md - Key article analysis
See Also
security-code-auditor — General security audits
privacy-first — Email/personal data prevention
Rationalizations
| Rationalization | Reality |
|---|
| "Small change, no logging." | AI interactions fall under transparency (Art. 50). |
| "Oversight later." | Oversight-by-design is mandatory for high-risk (Art. 14). |
Red Flags
Voice & Context
- Default:
professional + blog
- Reference:
voice-profiles skill for definitions and auto-detection.