Article 50 is not one generic notice requirement. It separates provider duties for AI systems that interact with people or generate synthetic content from deployer duties for emotion recognition, biometric categorisation, deepfakes, and certain public-interest text.
This page helps decide which disclosure belongs in the product interface, content workflow, model-output pipeline, workplace or public-facing notice, and which high-risk AI instructions for use must still be handled separately.
Structured answer sets in this page tree.
Cited legal and guidance references.
transparency duties apply from 2 August 2026. A limited grace period applies only to Article 50(2) marking and detection for systems placed on the market before that date: providers must comply from 2 December 2026. Content generated before 2 August 2026 does not require retroactive labelling. The final Code of Practice on Transparency of AI-Generated Content was published on 10 June 2026 as a voluntary implementation tool. On 8 July 2026 the Commission concluded that the Code adequately covers Articles 50(2), 50(4), and 50(5), and the AI Board adopted its assessment the next day. Adherence is not conclusive evidence of compliance. The Commission published final, non-binding Article 50 guidelines on 20 July 2026; the Regulation remains controlling law.
creates several distinct transparency duties. Providers must design direct-interaction AI systems so natural persons are informed that they are interacting with an AI system, unless that is obvious in context. Providers of AI systems, including general-purpose AI systems, that generate synthetic audio, image, video, or text content must ensure outputs are machine-readable and detectable as artificially generated or manipulated, subject to the Article 50 exceptions.
Deployers carry separate duties where their use exposes people to emotion recognition or biometric categorisation systems, creates or manipulates image, audio, or video content constituting a , or publishes AI-generated or manipulated text to inform the public on matters of public interest. also requires the information in paragraphs 1 to 4 to be clear, distinguishable, accessible, and given no later than the first interaction or exposure.
For a chatbot, voice assistant, agent, embedded support flow, or similar interface, first decide whether the system qualifies as AI, supports a genuine two-way exchange, interacts directly rather than through a human intermediary, and communicates with natural persons. Background systems, machine-to-machine communication, one-way data collection, and systems without direct contact fall outside this particular notice duty. If all four conditions are met, place the disclosure where the person first interacts with the system.
The notice can be unnecessary only when the AI nature is obvious from the viewpoint and context described in . That exception should be used narrowly in product review: record the interface, the user group, the context of use, and why the AI interaction would be obvious before relying on it.
separates provider-side technical marking from deployer-side public disclosure. Providers of systems that generate synthetic audio, image, video, or text content must ensure outputs are marked in a machine-readable format and detectable as artificially generated or manipulated, as far as technically feasible and taking account of content type, implementation cost, and the state of the art.
The provider marking duty does not apply where the AI system performs an assistive function for standard editing or does not substantially alter the input data or its semantics. It also has a law-enforcement exception where use is authorised by law to detect, prevent, investigate, or prosecute criminal offences. Record the function and output comparison before using either exception.
Deployers have a different obligation when they generate or manipulate image, audio, or video content constituting a : disclose that the content has been artificially generated or manipulated. For evidently artistic, creative, satirical, fictional, or analogous works, the obligation is limited to disclosure of the existence of generated or manipulated content in an appropriate way that does not hamper the display or enjoyment of the work. For AI-generated or manipulated text published to inform the public on matters of public interest, deployers must disclose the AI generation or manipulation unless 's law-enforcement or human-review/editorial-responsibility exceptions apply.
Deployers of emotion recognition systems or biometric categorisation systems must inform the natural persons exposed to the operation of the system, whether the operation occurs in real time or after the biometric data was captured. That disclosure duty is separate from data-protection duties: expressly states that personal data must be processed under the applicable EU data-protection instruments.
The notice record should therefore answer two questions at once. First, did the exposed person receive clear information about the operation of the system by the time of exposure? Second, did the privacy team assess the lawful basis, data categories, safeguards, and documentation required by the applicable data-protection regime?
transparency duties operate alongside Chapter III high-risk AI requirements. Article 50(6) says paragraphs 1 to 4 do not affect Chapter III obligations, while Article 13 requires high-risk AI systems to be sufficiently transparent for deployers and accompanied by concise, complete, correct, clear, relevant, accessible, and comprehensible instructions for use.
That means the same AI product can need two different transparency artifacts: a user-facing notice or label, and a deployer-facing Article 13 instruction set. Article 26 then requires deployers of high-risk AI systems to use the system in accordance with the instructions for use, assign competent human oversight, monitor operation based on those instructions, and keep logs where the logs are under their control.
This checklist covers product release, content publication, procurement onboarding, or deployment approval when may apply. The goal is to prove that the correct actor delivered the correct notice, label, or technical marking to the correct audience at the correct moment.
Do not use this checklist as a substitute for high-risk classification, prohibited-practice screening, privacy review, or sector-specific disclosure rules. transparency can sit beside those duties, and Article 50(6) preserves other transparency obligations under Union or national law.
No. covers user-facing transparency for direct interaction with an AI system. If the system is high-risk, Article 13 instructions for use and Article 26 deployer operating duties still need their own evidence and controls.
The deployer of the AI system that generates or manipulates image, audio, or video content constituting a must disclose that the content was artificially generated or manipulated, subject to the exceptions and special treatment for evidently artistic, creative, satirical, fictional, or analogous works.
Providers should keep the marking design, machine-readable format decision, detectability testing, content-type limitations, robustness and interoperability assumptions, release approvals, and records showing why any exception was or was not used.
Deployers should inform the natural persons exposed to the operation of the emotion recognition or by the time of exposure, and keep the notice aligned with applicable EU personal-data compliance records.
No. The limited grace period applies only to the (2) marking and detection duty for systems placed on the market before 2 August 2026; providers of those systems must comply from 2 December 2026. Interaction notices, exposed-person notices, and deployer content disclosures do not receive that grace period. Content generated before 2 August 2026 does not need retroactive labelling, although the Commission encourages it where possible.
Sorena can help separate provider marking duties, deployer disclosure duties, high-risk instructions for use, and privacy evidence into a practical Article 50 review record.
Ask questions tied to cited sources about Article 50 notices, AI-generated content labels, deepfake disclosures, and high-risk AI instruction handoffs.
Review your Article 50 trigger analysis, notice wording, labeling evidence, and high-risk AI transparency handoffs with Sorena.