EU AI Act Article 50: labels for deepfakes and AI-generated content

September 27, 2026

Since August 2, 2026, EU rules require AI-generated media to carry machine-readable marks and deepfakes to be clearly disclosed.

Synthetic MediaDeepfakesLegislationPlatformsDevelopersBrands

The short version

  • Article 50 of the EU AI Act requires providers of generative AI to mark outputs in a machine-readable way, and deployers to disclose deepfakes.
  • The obligations apply from August 2, 2026. AI systems already on the market before that date have until December 2, 2026 to meet the marking rule.
  • A voluntary Code of Practice, published June 10, 2026, explains how companies can comply.

What the law does

Article 50 of Regulation (EU) 2024/1689 sets four main transparency duties:

  • AI that talks to people: Providers must design systems so people know they are interacting with AI, unless it is obvious.
  • Marking AI output: Providers of systems that generate synthetic audio, images, video or text must mark outputs in a machine-readable format so they are detectable as AI-generated, as far as technically feasible.
  • Emotion recognition and biometric categorization: Deployers must inform the people exposed to them.
  • Deepfakes and public-interest text: Deployers who create deepfakes (realistic image, audio or video content resembling real people, places or events) must disclose that the content is artificially generated or manipulated. The same applies to AI text published to inform the public on matters of public interest, unless it has had human review and someone holds editorial responsibility.

For evidently artistic, creative, satirical or fictional works, the deepfake disclosure can be done in a way that doesn't spoil the work.

The AI Omnibus (Regulation (EU) 2026/1744), in force since July 27, 2026, added the grace period to December 2, 2026 for generative systems placed on the market before August 2, 2026.

Who it protects

People in the EU who see or hear AI-generated content, including people depicted in deepfakes.

What it means for you

Creators: Deepfakes of you shown in the EU must be labeled as AI-generated, whoever made them. A label doesn't make an unauthorized deepfake lawful under other laws.

Brands and agencies: Synthetic ads using realistic people need a clear disclosure in the EU.

Developers and platforms: Plan for watermarking or metadata and for labeling tools. The Code suggests using more than one marking technique.

What it doesn't do

  • It doesn't give you a right to consent to, or stop, a deepfake of you. It requires disclosure.
  • It doesn't create an individual damages claim. Enforcement is by national market surveillance authorities, with fines for non-compliance.
  • Some law-enforcement uses authorized by law are exempt.

Status and key dates

  • August 1, 2024: AI Act entered into force.
  • June 10, 2026: Final Code of Practice on Transparency of AI-Generated Content published.
  • July 27, 2026: AI Omnibus in force.
  • August 2, 2026: Article 50 obligations apply.
  • December 2, 2026: Marking deadline for generative systems already on the market before August 2, 2026.

Frequently asked questions

Do deepfakes have to be labeled in the EU? Yes. Since August 2, 2026, deployers who generate or manipulate deepfake content must disclose that it is AI-generated or manipulated, with a lighter approach for evidently artistic or satirical works.

Does Article 50 require watermarks? It requires machine-readable marking that is detectable as AI-generated, as far as technically feasible. Watermarking and metadata are common methods.

What you can do

  • Report unlabeled deepfakes of you to the platform hosting them, and keep copies.
  • Combine this with other remedies, such as GDPR rights or national personality rights.

Royall helps you keep a record of what you have and haven't consented to.

Sources

Last verified: 2026-09-27

This is general information, not legal advice.

onwards,
The Royall team