EU AI Act Transparency Rules Now in Force: What to Do
The EU AI Act transparency rules have applied since August 2026. Here is what a business with a chatbot or AI-made content should check, in plain words.
rabbitclip teamPublished: 6 min read
Short answer
The transparency rules of the EU AI Act have applied since 2 August 2026. According to the European Commission and the summary of Article 50, a system that interacts directly with people (a chatbot, for instance) must tell the person that it is AI; synthetic audio, images, video and text should be marked in a machine-readable way as far as is technically possible; and deepfake content must be disclosed. Whether a business outside the EU, a Turkish one for example, falls within scope depends on whether it serves users in the EU.
This article is not legal advice; ask your own lawyer about your situation. We simply relay what the official sources say, and we separate our own interpretation by marking it as our reading.
Below you will find when the rules began, what Article 50 asks for, how a business based in Turkey stands, what you can check on a site, which high-risk rules were postponed, and the mistakes made most often.
When did the AI Act transparency rules begin?
According to the European Commission's digital strategy page, the Act entered into force on 1 August 2024 and has been applied in steps. Prohibited practices and AI literacy came on 2 February 2025, and governance and general-purpose model obligations on 2 August 2025. On 2 August 2026 the Act began to apply generally, and the transparency rules have been in force since August 2026.
The AI Office and the authorities of the member states are responsible for enforcement. Our reading: the first day of the rules is behind us, so this topic no longer belongs on a list for later. That does not mean every site must change something at once; first you need to see whether the rule touches your business.
What exactly does Article 50 ask for?
Under the summary of Article 50, four things are expected from 2 August 2026. The information must be given at the first interaction, in a clear and accessible way. For the chatbot notice the summary also says it applies unless it is obvious to a reasonable person, so if it is already plain that the other side is AI, a separate notice may not be needed.
The full text of the Act is long and has exceptions. The list below gives only the outline of the summary; your lawyer should read the actual provision that applies to your system.
- Systems that interact directly with people must tell them that it is AI
- Output of systems producing synthetic audio, images, video and text should be marked in a machine-readable way as far as technically possible
- Deepfake content must be presented with a disclosure that it was generated or manipulated
- AI-generated text published on matters of public interest must be disclosed, with an exception where a human takes editorial responsibility
Does a business based in Turkey fall within scope?
The test in the sources is less where the business is established than where its users are. Whether a Turkish company falls in scope depends on whether it serves users in the EU. A business that serves only Turkey and does not target EU visitors, even if some arrive, is in a different position from an online shop that sells to Europe.
Our reading: if you sit on the border, first answer the question inside your own team: do we serve the EU? Pages in a European language, prices in euros and shipping to EU countries help answer it, but a lawyer gives the final word.
What should you check if your site has a chatbot or AI content?
What follows is not an official compliance list but a plain check. First make an inventory of everything produced by AI on your site and in your adverts: a chatbot, voice cloning, generated images and video, AI-written text.
Then put two questions next to each item: does the user understand at the first moment that it is AI, and is it possible to mark or disclose the output.
- State in the first message of the chatbot that it is AI
- Check the marking options of the tool you use for generated images and video
- On public-interest news and information texts, state the role of AI and the editorial responsibility
- Say clearly when generated content resembles a real person
Which rules did the Digital Omnibus postpone?
According to the European Commission, the AI part of the Digital Omnibus was proposed on 19 November 2025, reached political agreement on 7 May 2026 and entered into force on 27 July 2026. The rules for high-risk systems in areas such as biometrics, education and employment were postponed to 2 December 2027. High-risk systems embedded in products were postponed to 2 August 2028.
A new prohibition will also apply in December 2026: AI that produces non-consensual sexual content and child abuse material. Our reading: the delay is for high-risk systems, and the transparency rules stayed in force in August 2026. If you have heard of a separate transition period for marking, we could not confirm it in our sources; we give no date, so ask your lawyer.
What mistakes are common, and who can skip this article?
The most common mistake is saying it does not concern us without reading the scope. The second is the opposite: trying to stamp every sentence on a site as AI. The rule is for certain systems and certain content; labelling everything without knowing the scope is both unnecessary and potentially misleading.
If your site has no chatbot, no generated images and no AI-written text, and you do not serve users in the EU, this topic may not be on your agenda today. Still, when you add a service, such as a new bot or a generative tool, remember to ask the question.
- Saying it does not concern us without reading the scope
- Putting an AI label on everything in bulk
- Not knowing what a third-party tool produces
- Trusting a blog post instead of legal advice
In short: the transparency rules have applied since August 2026, the dates for high-risk systems were postponed, and a Turkish business is in or out of scope depending on whether it serves users in the EU. The next step is to list the AI on your site and take the questions to your lawyer; this article is not legal advice. If you would like to draw up that inventory together from the technical side, we are glad to hold a calm discovery conversation at rabbitclip.
FAQ
Must I say that my chatbot is AI?
Under the Article 50 summary, systems that interact directly with people must disclose it, but where it is already obvious to the person a notice may not be needed. Your lawyer can say whether the rule applies to you. In practice one sentence in the first message is a simple way.
Do I need to label a blog post I wrote with AI?
According to the summary, only AI-generated text published on matters of public interest must be disclosed, with an exception where a human takes editorial responsibility. Not every blog post falls under it. Ask your lawyer about your own content.
My site is only in Turkish. Am I still in scope?
The test is whether you serve users in the EU. If you serve only Turkey, your position is different. If you are on the border, ask your lawyer.
When do the high-risk system rules start?
Under the Digital Omnibus, systems in areas such as biometrics, education and employment were postponed to 2 December 2027, and those embedded in products to 2 August 2028. The transparency rules were not postponed.
Sources
The specific facts in this article rest on these official pages; check them for the current position.
