Article 50 of the EU AI Act requires you to tell people when they are dealing with AI and to mark AI-generated content. It applies from 2 August 2026, it was not extended by the Digital Omnibus, and it reaches ordinary organisations — chatbots, synthetic media, AI-written content — not just AI developers. Here is what has to be disclosed, and by when.
Most of the EU AI Act attention in Ireland has gone to two things: whether your AI is high-risk, and when the deadlines bite. Article 50 sits outside both conversations, and that is exactly why it catches organisations out. It is not risk-tiered — it applies regardless of how risky your system is — and its deadline did not move in the Digital Omnibus. From 2 August 2026, transparency is a live obligation for ordinary Irish organisations, not just AI vendors.
What Article 50 actually requires
Article 50 is the AI Act's transparency layer. It covers four situations, and the common thread is simple: people should know when they are dealing with AI, and content should carry an honest signal of how it was made.
- AI systems that interact with people must be designed so the person is informed they are interacting with AI — unless it is obvious from the context. In practice this is your chatbot, your voice agent, your AI-driven support tool.
- Synthetic audio, image, video and text generated or manipulated by AI must be marked in a machine-readable format as artificially generated.
- Deepfakes — AI-generated or manipulated image, audio or video that resembles real people, places or events — must be disclosed as such.
- AI-generated text published to inform the public on matters of public interest must be disclosed as AI-generated, unless a human has reviewed it and someone holds editorial responsibility.
The obligation falls on both providers (who build the systems) and deployers (who use them). Most Irish organisations are deployers — and deployers are squarely in scope.
Who this reaches that does not expect to be reached
The organisations most likely to be caught off guard are not technology companies. They are:
- Any business running a customer-facing chatbot or AI assistant on its website or in its app.
- Marketing, communications and content teams publishing AI-generated or AI-edited images, video or copy.
- Professional services and public bodies producing public-facing content where AI drafted material without clear human editorial sign-off.
None of these are high-risk uses. All of them touch Article 50. The transparency obligation is decoupled from the risk classification precisely so that it applies broadly.
The timeline — and the one piece of breathing room
Article 50 applies from 2 August 2026, in step with the AI Office of Ireland becoming operational. The Digital Omnibus that extended the high-risk deadlines to December 2027 did not extend Article 50.
The one concession: the machine-readable marking of AI-generated content by systems already on the market gets a grace period to 2 December 2026. That is a narrow technical allowance for existing systems to implement marking — not a deferral of the disclosure obligations, and not a reason to wait.
For where Article 50 sits in the wider enforcement picture, and what the AI Office does from day one, see our guide to the AI Office of Ireland.
What to do before August
Transparency is one of the more tractable parts of the AI Act — it is largely about disclosure discipline, not conformity assessments. A workable pass:
- Inventory every point where AI touches a person or produces public-facing content. Chatbots, assistants, generated media, AI-drafted copy. You cannot disclose what you have not mapped.
- Add clear AI-interaction notices wherever a person could reasonably think they are dealing with a human and it is not otherwise obvious.
- Set a content-marking and labelling standard for AI-generated media and deepfake-style content, and check whether your tools support machine-readable marking.
- Define editorial responsibility for AI-assisted public content, so the human-review exemption actually applies to you rather than being assumed.
- Write it down. A short transparency policy with a named owner is what turns "we generally disclose" into something you can evidence to a regulator.
This work also feeds the broader picture — an AI inventory built for Article 50 is the same inventory you need for risk classification and the AI literacy obligation.
Article 50 is small in the text and wide in its reach. If you want a clear read on where your organisation stands — what has to be disclosed, where, and by when — Acuity AI Advisory's fixed-fee, vendor-neutral EU AI Act readiness review covers transparency alongside inventory, risk classification and a remediation roadmap. For ongoing support, we provide independent EU AI Act consulting for Irish organisations.