Most of the EU AI Act's early coverage focused on prohibited practices and high-risk systems. For marketing teams, Article 50 is the part that matters in day-to-day work. It became applicable on August 2, 2026 and covers disclosure: telling people when they are talking to AI, and labelling content that AI made. This page summarizes what it requires. It is a factual summary, not legal advice.
The Four Obligations
| Paragraph | Who | Requirement |
|---|---|---|
| 50(1) | Providers | AI systems that interact with people (chatbots, voice assistants, agents) must make it clear the person is dealing with AI, unless that is obvious to a reasonably well-informed person |
| 50(2) | Providers | Systems that generate synthetic audio, image, video or text must mark outputs as AI-generated in a machine-readable, detectable format. Assistive editing that doesn't substantially change the input is exempt |
| 50(3) | Deployers | Anyone using emotion recognition or biometric categorisation must inform the people exposed to it |
| 50(4) | Deployers | Deepfakes must be disclosed as artificially generated. AI-generated text published to inform the public on matters of public interest must be disclosed, unless it has gone through human review and someone holds editorial responsibility |
For artistic, satirical or fictional work, the deepfake disclosure can be light, enough to note the manipulation "in a manner that does not hamper the display or enjoyment of the work."
The Dates
| Date | What happens |
|---|---|
| June to July 2026 | The Commission finalizes the Code of Practice on Transparency of AI-Generated Content and adopts Article 50 guidelines. Sources give dates between June and July 20 |
| August 2, 2026 | Article 50 obligations apply |
| December 2, 2026 | End of the grace period for machine-readable marking by generative AI systems that were on the market before August 2 |
| February 2, 2027 | Interoperability requirement for watermark detection, as reported by Paul, Weiss under the Digital Omnibus changes |
Content published before August 2, 2026 doesn't need to be labelled after the fact.
Penalties and Enforcement
Breaching Article 50 can bring fines of up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher, with proportionality for SMEs. National market surveillance authorities enforce, and the AI Office covers systems under its direct supervision. The EU AI Act enforcement tracker follows actions as they happen.
The Code of Practice and the EU Icons
Signing the voluntary Code of Practice is the most direct way for providers to show they meet the marking and detection rules. Organizations that don't sign have to justify their own measures to regulators. Paul, Weiss counted about 190 signatories by July 31, 2026. The Commission also published standard icons for deployers: one for fully AI-generated content, one for AI-modified content, and a basic icon with an optional interactive second layer.
What It Means for Marketing Teams
- Chatbots and sales agents on your site that talk to EU users need a clear AI disclosure at first contact.
- Synthetic images, video and voice in campaigns carry provider-side machine-readable marks. Deepfake-style creative, such as a realistic person who doesn't exist or a synthetic spokesperson, also needs a visible disclosure from you as the deployer.
- AI-drafted articles need disclosure only when they inform the public on matters of public interest and no human has taken editorial responsibility. Product copy with human review is generally outside 50(4).
Brand Visibility Implications
Machine-readable marking means AI systems will increasingly be able to tell AI-generated content from human-made content at crawl time. No major assistant has said it down-ranks marked content. Still, answer engines already prefer sources with clear authorship and editorial accountability, and Article 50 turns that accountability into something they can detect. Brands that publish human-reviewed, named-author content are in the stronger position. See the AI policy tracker for other jurisdictions.
Methodology
Figures on this page are drawn from primary announcements and documentation wherever they exist, with established trade and legal press used only to corroborate or to fill gaps a primary source leaves open. Sources: Article 50 text, European Commission quick facts, Cooley, Paul, Weiss. Where outlets disagree on a date or number, the page gives the range or names the disagreement rather than choosing one. Status as of September 2026; this area changes monthly, so re-check before quoting.
How Presenc AI Helps
Presenc AI tracks how AI assistants describe your brand across markets, including EU prompts and languages. Teams can see whether their human-reviewed content is the source answers draw on, and spot when AI-generated third-party content starts shaping those answers instead.