Article 50 of the EU AI Act makes AI-generated content labeling a legal duty — transparency obligations have applied since August 2, 2026, and the machine-readable marking phase-in for pre-existing systems ends December 2, 2026 (further phases: December 2, 2027 and August 2, 2028). Take the 6-question self-test below and get a compliance risk readout citing the exact Article 50 limb behind every gap.
Article 50 is the EU AI Act’s transparency rule: providers and deployers must disclose AI-generated content in a machine-readable way. Transparency obligations have applied since August 2, 2026.
Countdown to December 2, 2026 (CET, Brussels). Want the embeddable widget version? EU AI Act Countdown widgets.
Answer for your AI-generated ads and content that reaches audiences in the EU.
1. Do you run AI-generated or AI-manipulated ads — images, video, or audio a viewer could mistake for real?
2. Is the AI origin disclosed on the ad itself, at first exposure — not just in metadata or a buried caption?
3. Do any ads use AI-generated faces, voices, or virtual influencers (deepfakes)?
4. Is your AI content machine-readable-marked (watermark or C2PA-style provenance)?
5. Do you publish AI-generated text to inform the public (news, health, policy, environment) without genuine human editorial review?
6. Do you run AI chatbots or AI customer agents without clearly telling users they are talking to AI?
Transparency breaches carry fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher (the AI Act's steepest band — up to €35M / 7% — applies to prohibited practices under Article 5). This readout is automated guidance, not legal advice.
Some do. Visible disclosure becomes particularly relevant when an AI-generated or manipulated advertisement realistically depicts a person, object, product, place, entity, or event and could falsely appear authentic or truthful. Advertising content is excluded from the lighter disclosure regime for genuinely artistic or satirical work.
No. Standard editing tasks fall outside the marking duty, including grammar correction, spellchecking, minor stylistic polish, and AI-generated translations. Summaries and substantive rewrites still require marking. Visible disclosure is triggered by deepfake content and by unreviewed public-interest text.
Breaching the Article 50 transparency obligations carries fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher. The AI Act's steepest band — up to €35 million or 7% of global turnover — applies to prohibited AI practices under Article 5.
Purely promotional copy may sit outside the disclosure rules, but content that informs the public on regulatory, technical, health or environmental matters is more likely to be flagged. Genuine human authorship and editorial ownership are the cleanest way to stay on the right side of the line.
Yes, they can. The Act follows the AI output rather than the company's location: if your AI system's output is used in the European Union — for example AI-generated campaign assets reaching EU audiences — you can be in scope wherever you are established.