The date
2 August 2026. That is the general application date set by Article 113, and Chapter IV, which holds the transparency rules, follows it.
The rule
An AI system built to interact with people has to be designed so those people know they are dealing with an AI, unless it is already obvious.
The reach
It is not only an EU problem. The regulation follows the output, so a coach in London or Denver with clients in the EU can be caught by it.
This article explains what the regulation says and cites the articles so you can read them yourself. It is general information, not legal advice, and it does not create a lawyer and client relationship. If you have an EU audience and money riding on the answer, take proper advice.
What changes on 2 August 2026
One article, four obligations, and a carve-out you should not lean on.
The EU AI Act applies in stages. The prohibited practices in Chapters I and II came first, in February 2025. Parts of the high risk classification rules come later, in August 2027. The transparency obligations that matter for anything that talks to your audience sit in Article 50, inside Chapter IV, and they follow the general application date. Article 113
Article 50(1) is the one that catches AI clones. It requires providers to ensure that AI systems intended to interact directly with people are designed and developed so that the people concerned are informed they are interacting with an AI system. There is an exception where that is obvious from the point of view of a person who is reasonably well informed, observant and circumspect. Article 50(1)
An AI clone is the one case where nothing about it is obvious. It uses a real person's name, knowledge and turns of phrase. That is the whole point of it, and it is exactly why the exception is a bad thing to rely on.
A generic support bot on a checkout page might well be obvious. A conversation that opens in your voice, answers with your frameworks and signs off the way you do is not. If a reasonable person could finish that conversation still believing they had spoken to you, you are not inside the exception.
Does it apply to you?
The most common misreading is that this is somebody else's jurisdiction.
Article 2 sets the scope, and it is written to follow effects rather than addresses. It covers providers placing AI systems on the Union market irrespective of whether they are established in the Union or in a third country. It also covers providers and deployers located in a third country where the output produced by the AI system is used in the Union. Article 2(1)(a), Article 2(1)(c)
| Your situation | Assume you are in scope? |
|---|---|
| You are in the EU and your clone talks to anyone | Yes |
| You are outside the EU, and EU clients use your clone | Yes, the output is used in the Union |
| You sell a course worldwide and cannot control who buys | Treat as yes unless you geo-block the EU |
| Your clone is private, internal to your team, no EU staff | Lower risk, but check with an adviser |
Provider or deployer?
The regulation splits duties between the two. A provider develops an AI system and places it on the market or puts it into service under its own name or trademark. A deployer uses an AI system under its own authority, outside of purely personal activity. Article 3(3), Article 3(4) The Article 50(1) duty to design the system so users know they are dealing with an AI sits with the provider. The deepfake disclosure duty in Article 50(4) sits with the deployer.
For most coaches the platform is the provider and you are the deployer. But notice the words "under its own name or trademark" in the provider definition. If you take a clone to market as your own branded product, the line genuinely blurs, and which side you fall on is the sort of question worth putting to an adviser rather than guessing at.
Either way, you cannot outsource the part that matters. You are the one whose face and name are attached to the output, and you are the one your audience will hold responsible.
What you have to disclose
Four obligations sit in Article 50. Not all of them will apply to you.
Tell people it is an AI
The system has to be designed so people are informed they are interacting with an AI system. This is the obligation that applies to essentially every AI clone, and it is on the provider. Article 50(1)
Mark synthetic output as machine readable
Providers of systems that generate synthetic audio, image, video or text must ensure the outputs are marked in a machine readable format and detectable as artificially generated or manipulated. The regulation qualifies this by what is technically feasible, and exempts assistive editing that does not substantially alter the input. Article 50(2)
Disclose deepfakes you publish
Deployers of a system producing deepfake image, audio or video must disclose that the content has been artificially generated or manipulated. There is an exception for work that is evidently artistic, creative, satirical or fictional. If you publish a video avatar of yourself, this is your obligation, not your platform's. Article 50(4)
Disclose AI text on public interest matters
Where AI generated text is published to inform the public on matters of public interest, deployers must disclose it. This one has a carve-out that matters: it does not apply where the content has undergone human review or editorial control and a person holds editorial responsibility for it. Article 50(4)
Read that last carve-out again if you publish AI assisted articles. Editing and standing behind what goes out is not a loophole, it is the standard the regulation is pointing at.
What happens if you ignore it
The number quoted everywhere is the enterprise number. It is not the one that applies to most readers of this page.
Non compliance with Article 50 falls under Article 99(4)(g). That paragraph sets fines of up to 15,000,000 EUR or, if the offender is an undertaking, up to 3% of total worldwide annual turnover for the preceding financial year, whichever is higher. Article 99(4)
For SMEs and startups, Article 99(6) flips it. The cap becomes whichever figure is lower, not higher.
That is a meaningful difference, and most coverage of the AI Act skips it because most coverage is written for corporate counsel. If you are a solo coach or a small studio, the headline 15 million figure is not your exposure. It is still not a reason to do nothing, because the reputational cost of being caught passing an AI off as a person lands long before any regulator does.
How to comply
Most of this is a morning's work, and none of it needs a lawyer to start.
Name it in the first message. The obligation bites at the point of interaction, so the opening line is where it belongs. Something like "Hi, I am the AI version of Lucy, trained on her coaching material" does the job in one sentence.
Make the name honest. "AI Lucy" tells the truth before anyone reads a word. A clone presented under the bare human name is the version that gets you in trouble.
Put it in the interface, not the policy page. A line in your terms is not informing someone at the point they interact. Label the chat window itself.
Give people a way to reach a human. Not strictly an Article 50 requirement, but it is what turns a disclosure from a warning into a service.
Label any video or voice avatar you publish. That is the Article 50(4) deployer duty and it sits with you, wherever you built the avatar.
Ask your platform what it marks. The machine readable marking duty is the provider's. You cannot implement it yourself, so ask, and get the answer in writing.
Why disclosure is the way forward
The regulation is the deadline. It is not really the reason.
It is tempting to read Article 50 as a tax on something that was working fine. That reading gets the economics backwards. An AI clone is worth something because a specific person stands behind it. Strip out the fact that a real expert trained it, curated it and takes responsibility for it, and what is left is a chatbot with a stolen voice.
Concealment is also structurally fragile. A clone that is passing as human has to keep passing forever, in every conversation, with every client, and one screenshot ends it. A clone that says what it is has nothing left to be exposed. It can be linked to, quoted, embedded and sold without a trapdoor underneath it.
The clones that survive the next few years will be the ones that were never pretending in the first place.
There is a practical version of this too. Disclosure sets expectations, and expectations are what protect you when the clone gets something wrong. A client who knows they are talking to your AI treats a bad answer as a bug. A client who thought they were talking to you treats the same answer as a betrayal. Same output, completely different consequence, and the only difference is one sentence at the start of the conversation.
If you are still deciding how to introduce a clone to your audience, our guide on whether an AI coach hurts your reputation covers the trust side, and what an AI twin actually is covers the mechanics.
Frequently Asked Questions
The questions people ask once the date lands in their calendar.
When do the EU AI Act transparency rules start applying?+
Article 50 sits in Chapter IV of the EU AI Act, which follows the regulation's general application date of 2 August 2026. Article 113 sets that date, with earlier dates for the prohibitions in Chapters I and II and a later date of 2 August 2027 for parts of the high risk classification rules.
Do I have to label my AI clone as AI?+
Article 50(1) requires that AI systems intended to interact directly with people are designed so that those people are informed they are interacting with an AI system, unless that is obvious to a reasonably well informed, observant and circumspect person. An AI clone built to sound like a specific named human is the case where it is least obvious, so disclosure is the safe assumption.
Does the EU AI Act apply to me if I am based in the UK or the US?+
It can. Article 2 applies the regulation to providers placing AI systems on the Union market irrespective of whether they are established in the Union or in a third country, and to providers and deployers located in a third country where the output produced by the AI system is used in the Union. A coach outside the EU whose clone answers clients in the EU should assume they are in scope.
What are the fines for breaching the AI Act transparency rules?+
Non compliance with Article 50 falls under Article 99(4)(g), which sets fines of up to 15,000,000 EUR or, for an undertaking, up to 3% of total worldwide annual turnover for the preceding financial year, whichever is higher. Article 99(6) reduces this for SMEs including startups, where the cap is whichever of the relevant figures is lower.
Where should the AI disclosure appear?+
Article 50(1) requires people to be informed at the point they interact with the system, so the disclosure belongs in the first thing the user sees rather than in a policy page. In practice that means naming the assistant as AI in its opening message and in the interface around it, so nobody has to go looking to find out who they are talking to.
Build a clone you can put your name on
An AI version of you that answers in your voice, in 100+ languages, and says what it is from the first message. Start on the free tier and see how it handles your real questions.
Free tier available. No credit card needed.