Art. 50 of the AI Act governs the transparency and labeling obligations for AI-generated content and has applied since August 2, 2026. This article explains what that means for you as a deployer.
Providers and deployers have different obligations
Art. 50 addresses providers and deployers with different obligations:
- Provider (Eye-Able): a duty to disclose direct AI interaction (para. 1, for example with the assistant) and machine-readable labeling of synthetic content (para. 2). Eye-Able fulfills these obligations on the product and provider side.
- Deployer (you as the customer): disclosure in cases of emotion recognition or biometric categorization (para. 3), as well as for deepfakes and for AI-generated or AI-manipulated text on matters of public interest (para. 4).
What does this mean for you specifically?
Using the Eye-Able software solutions generally creates no independent need for action on your part in the form of visible labeling:
- No deepfakes, no emotion or biometric recognition: The Eye-Able functions do not create deepfakes and do not perform emotion recognition or biometric categorization. The deployer obligations under Art. 50 para. 3 and para. 4 are therefore not applicable.
- No newly generated public AI text under your responsibility: The functions assist in processing your own, human-authored content. The editorial responsibility for the source content remains with you, so the exception for content under human editorial control applies, provided the AI-processed output is reviewed by you before publication.
- Machine-readable labeling is the provider's responsibility: Where Art. 50 para. 2 applies, the technical, machine-readable labeling is Eye-Able's responsibility as the provider. You do not need to take any technical measures of your own on your website.
If you additionally use other generative AI to create your own editorial content, this must be assessed separately by you.
Voluntary recommendation
A short, voluntary transparency notice — for example in your accessibility statement or privacy policy — stating that AI-supported accessibility functions are in use builds trust. This is not a legal obligation. Eye-Able is happy to provide suggested wording.
Deadlines and responsibility
The transparency obligations under Art. 50 para. 1, 3, and 4 have applied since August 2, 2026. For the provider-side machine-readable labeling under para. 2, a transition period until December 2, 2026 applies to systems placed on the market before August 2, 2026. The frequently cited date of December 2, 2026, therefore does not mark the start of a new obligation but the end of this transition period.
Important: Each EU member state designates its own market surveillance authority under the AI Act. You can find your country's designated authority here.
Further articles
In the following article, you’ll find further information on similar topics:
- The EU AI Act and the Eye-Able software solutions
- Classification of Eye-Able's AI functions under Art. 50 of the AI Act
If you have any further questions, our team will be happy to help! In the following article, you’ll find out how to get in touch with us: