EU AI Act — article-by-article guides

Provider AND deployer training programmes for staff who use, oversee, or are affected by AI. Already binding.

Continuous, iterative risk identification + analysis + mitigation across the lifecycle.

Relevant, representative, error-free training/validation/test datasets with documented bias examination.

Provider must supply instructions for use containing all info deployers need to use the system safely + lawfully.

Effective natural-person intervention, override, and stop capabilities for high-risk AI.

Technical guarantees + adversarial-attack resilience + audit-grade logging for high-risk AI.

Operational obligations on companies USING high-risk AI, plus the Fundamental Rights Impact Assessment for public-sector + insurance deployers.

Self-assessment (Annex VI) or Notified Body audit (Annex VII) before placing on EU market. EU declaration of conformity + CE marking.

Machine-readable AI-content marking + visible deepfake disclosure. C2PA Content Credentials + SynthID-class watermark.

Documented PMM plan + continuous data collection + feedback loop into Article 9 RMS + Article 73 incident reporting.