Who this applies to
This obligation applies to
providers of AI systems that interact with natural persons (e.g., chatbots) or generate synthetic content (e.g., deepfakes, AI-generated text/image/audio). It also covers
deployers of such systems when they fall under high-risk classifications or general-purpose AI (GPAI) models with systemic risk. Scope is defined under
Article 50(1) (transparency for AI systems) and
Article 52(1)-(3) (obligations for GPAI providers).
What is required
- Disclosure of AI interaction: Systems designed to interact with natural persons must inform users they are engaging with an AI, unless obvious from the circumstances (Article 50(1)).
- Synthetic content labeling: AI-generated or manipulated content (e.g., deepfakes, text, audio, video) must be clearly marked as artificially generated or manipulated, including via machine-readable metadata where technically feasible (Article 50(2)).
- GPAI transparency documentation: Providers of general-purpose AI models must publish detailed summaries of the content used for training, unless prohibited by copyright or trade secrecy (Article 52(1)).
- Systemic risk disclosures: Providers of GPAI models with systemic risk must additionally disclose:
- Energy consumption metrics (
Article 52(2)(a)).
- Data governance measures, including bias mitigation and compliance with EU fundamental rights (
Article 52(2)(b)).
- Risk assessment reports for downstream deployers (
Article 52(3)).
Key deadlines
The primary deadline for this obligation is
August 2, 2026.
Enforcement patterns
AI Act enforcement begins August 2, 2026. No precedent currently exists. This page will be updated as enforcement cases emerge.
Cross-border considerations
Data is insufficient to identify jurisdiction-specific implementation patterns for transparency obligations under the AI Act. No member state authorities or national deviations are cited in the provided enforcement or cross-reference materials.