When Article 50 transparency obligations apply, who must inform users, and what changed after the Omnibus.
Upcoming
Users must be informed when interacting with an AI system; AI-generated or manipulated content, including deepfakes, must be disclosed; disclosure duties apply for emotion-recognition and biometric-categorization systems. Article 49 EU database registration framework and governance/enforcement provisions also apply. This date was not deferred.
Changed by Omnibus Original: August 2, 2026
The machine-readable marking/watermarking duty under Article 50(2) applies to generative systems placed on the market before Aug 2, 2026 after a four-month grace period. Systems placed on the market on or after Aug 2, 2026 must comply from placement.
Changed by the Digital Omnibus on AI (Regulation (EU) 2026/1744), in force since July 27, 2026.
Article 50 is about telling people when AI is materially involved in what they are seeing, hearing, or interacting with. The deadline data covers direct AI interaction notices, disclosures for AI-generated or manipulated content including deepfakes, and disclosures for emotion-recognition and biometric-categorization systems. It also separates the main transparency date from the machine-readable marking grace period for older generative systems.
In product terms, this means labels, notices, interface copy, and content workflows. A user should not have to guess whether they are dealing with an AI system. A viewer should not be left thinking synthetic or manipulated content is ordinary human-made media when the disclosure duty is triggered. For generative systems, the machine-readable marking question belongs in the product and engineering backlog, not only in legal review.
A customer-support tool that lets users interact with an AI assistant is a simple example. A marketing or media platform that creates synthetic image, audio, video, or text can also trigger review. A workplace or education product using emotion recognition or biometric categorization deserves especially careful scoping because the deadline data calls out those systems directly.
The main misunderstanding is treating transparency as a privacy-policy update. A privacy page can help, but it does not replace a clear in-context disclosure. Another mistake is assuming the Omnibus delayed all transparency work. The deadline data says the main Article 50 transparency date was not deferred, while the machine-readable marking grace period for older generative systems changed.
Map every user-facing AI interaction and every workflow that produces synthetic or manipulated content. For each one, decide where the disclosure appears, who sees it, and whether the wording is understandable without legal training. For generative systems, separate systems already placed on the market before the relevant threshold from systems placed later, because the deadline data treats those differently. If the product involves emotion recognition or biometric categorization, escalate it before launch rather than treating it as ordinary analytics.
