Short answer

Since 2 August 2026, a chatbot or voice bot that talks to people in the EU has to tell them they are talking to an AI, at the latest at the first interaction, unless that is already obvious. The rule is Article 50 of the EU AI Act. It also covers AI generated images, audio and video, deepfakes, and AI written text on matters of public interest. Breaking it can cost up to 15 million euros or 3% of worldwide turnover, and for small companies the lower of the two applies.

2 Aug 2026Article 50 applies to chatbots, deepfakes and AI news text
€15M or 3%Maximum fine, whichever is higher (the lower of the two for SMEs)
2 Dec 2026Marking deadline for generative AI placed on the market before August

Most small businesses meet this rule the moment they add a support widget, an AI sales assistant or an AI phone agent to their site. You do not need to have built the model yourself. If people in the EU can reach the bot, the rule reaches you too, even when your company is in the United States.

This guide covers which duty sits with whom, the exact wording you can put in front of users, where it has to appear, the few cases where you can skip it, how the new Code of Practice and the AI Omnibus change the picture, and a checklist you can finish in an afternoon. Every date and figure comes from the text of the AI Act or from the European Commission, listed under Sources.

What Article 50 requires

Article 50 has four duties. Two sit with the provider, the company that develops an AI system or has it developed and offers it under its own name. Two sit with the deployer, the business that uses an AI system in its work. Many small companies are both, for example when they build their own assistant on top of a model API and put it on their website.

DutyWhoWhat it means in practice
Tell people they are talking to AI (Art. 50(1))ProviderA chatbot, avatar or voice agent must be designed so users are informed they are dealing with an AI, unless that is obvious to a reasonably well informed person.
Mark AI generated media (Art. 50(2))ProviderSynthetic audio, images, video and text must carry a machine readable marker showing they were made or changed by AI.
Disclose emotion recognition or biometric categorisation (Art. 50(3))DeployerIf you run a system that reads emotions or sorts people by biometric traits, you must tell the people exposed to it.
Label deepfakes and AI news text (Art. 50(4))DeployerDeepfake images, audio or video must be disclosed. AI written text published to inform the public on matters of public interest must be disclosed, unless a human reviewed it and a person holds editorial responsibility.

Under Article 50(5) the information has to be given in a clear and distinguishable manner, at the latest at the time of the first interaction or exposure, and it has to meet accessibility requirements. The Commission adds that people must be able to understand and perceive it without any special technical tools, for example through a visible or audible label. A line on page nine of your Terms of Service does not do that.

Article 50 by the numbers

4Separate transparency duties in Article 50(1) to 50(4)AI Act, Art. 50
2 Aug 2026The date Article 50 started to apply across the EUAI Act, Art. 113
€15M or 3%Maximum fine for breaking a transparency duty, whichever is higherAI Act, Art. 99(4)
2 Dec 2026Marking deadline for generative AI placed on the market before 2 August 2026AI Omnibus
95 and 192Providers and deployers that signed the Code of Practice on AI generated content by 31 July 2026European Commission
27 Jul 2026The AI Omnibus, Regulation (EU) 2026/1744, entered into forceOfficial Journal

Provider or deployer: which one are you?

The duty you carry depends on your role for each AI system, not on the size of your company. These are the setups small businesses meet most often.

Your setupYour roleWhat you need to do
You add a ready made AI chat widget from a helpdesk vendor to your siteDeployer. The vendor is the provider.Check that the widget tells users it is an AI and keep that setting on. Do not give it a human name and photo that hide it.
You build your own assistant on a model API and run it under your brandProvider and deployerYou own the Article 50(1) disclosure. Build it into the first message and the chat header.
You run an AI phone agent that answers your business lineUsually provider and deployer, or deployer of a vendor's agentThe first sentence of the call must say it is an automated AI assistant.
You publish AI generated images or video of real looking people, places or eventsDeployerLabel them as AI generated where people see them (Article 50(4)).
You publish AI written articles on public interest topics with no editorial reviewDeployerLabel the text as AI generated, or put a person in charge of reviewing it before it goes out.
You use AI to draft product descriptions that your staff check before publishingDeployerNo label duty for that text. Product copy is not news on a matter of public interest, and it is reviewed by a person.

Does it apply to a business outside the EU?

Yes, if the output of your AI system is used in the EU. The AI Act reaches providers and deployers established outside the Union when their system's output is used inside it (Article 2). A US store whose support bot answers a shopper in Madrid is in scope for that conversation. The simplest way to stay safe is to show the disclosure to everyone rather than trying to detect where each visitor is.

Wording you can use today

There is no required sentence. The test is whether an ordinary person understands, at first contact, that they are dealing with an AI. These examples pass that test and fit most sites.

Website chat widget, first message

Hi, I'm the Outline Technologies assistant. I'm an AI, not a person. I can answer questions about orders and returns, or pass you to our team.

Voice or phone agent, first sentence

Hello, you're speaking with an automated AI assistant for Outline Technologies. You can ask for a human at any time.

AI generated image or video in a post or ad

This image was created with AI.

An article written with AI and published without editorial review

This article was written by an AI system and has not been reviewed by an editor.

Keep the label next to the thing it describes. A persistent "AI" tag in the chat header, plus the opening line, is better than a single sentence that scrolls away. In a voice product, say it out loud rather than relying on a note in an app.

Where the disclosure must appear

  • In the interface itself. The opening message of the chat, the first sentence of the call, or a caption on the image. Not only in a policy page.
  • Before or at the first interaction. Showing it after the user has already shared personal details is too late.
  • In a form people can perceive. Read it aloud in voice products. Make sure a screen reader announces it in chat widgets, since Article 50(5) requires accessible information.
  • Again in your legal pages. Your privacy policy should say which AI tools process chat messages. An AI transparency notice explains the rest in one place.

When you can skip it

The chatbot duty does not apply when it is obvious from the circumstances that the user is dealing with an AI. The Commission judges this from the point of view of an average person who is reasonably well informed, circumspect and observant, and says the exception must be read narrowly. A product called "AI Writer" whose whole purpose is generating text is an obvious case. A support chat that uses a human first name and a photo of a person is the opposite, and it is exactly what the rule targets.

The text duty for deployers does not apply when a human has reviewed the text and a person or company holds editorial responsibility for publishing it. The Commission is clear that a quick spelling check is not enough: the review has to examine the substance. The marking duty for providers does not cover tools that only assist with standard editing or do not substantially change what the user put in. Content generated before 2 August 2026 does not need to be labelled after the fact.

The Code of Practice on AI generated content

In June 2026 the EU AI Office published a voluntary Code of Practice on Transparency of AI generated Content. It explains how providers can mark AI output in a machine readable way and how deployers can label deepfakes and AI news text. On 8 July 2026 the Commission found that the code adequately covers Articles 50(2), 50(4) and 50(5), and the AI Board confirmed that assessment the next day.

Signing is optional, but it gives a company a recognised way to show it complies. By 31 July 2026, 95 companies had signed the provider section, including Anthropic, Google, Meta, Microsoft, Mistral and OpenAI, and 192 had signed the deployer section. About half of all signatories are small or recent companies. Businesses that do not sign must show compliance in other ways and can expect more questions from regulators.

What changed with the AI Omnibus

The EU's Digital Omnibus on AI, Regulation (EU) 2026/1744, was published in the Official Journal on 24 July 2026 and entered into force on 27 July 2026. It pushed back the main high risk obligations: rules for high risk systems listed in Annex III now apply from 2 December 2027, and rules for AI built into products covered by Annex I from 2 August 2028.

It did not move the chatbot duty or the deepfake duty, which apply from 2 August 2026. The one Article 50 change is a four month grace period for the machine readable marking duty in Article 50(2): generative AI systems placed on the market before 2 August 2026 have until 2 December 2026 to comply. The Omnibus also added a ban on AI systems that generate non consensual sexually explicit images or child sexual abuse material, which applies from the same December date.

Timeline

DateWhat applies
1 August 2024The AI Act enters into force
2 February 2025Bans on prohibited AI practices and the AI literacy duty
2 August 2025Rules for general purpose AI models, and the penalty rules
27 July 2026The AI Omnibus enters into force
2 August 2026Article 50 transparency duties for chatbots, deepfakes and AI news text
2 December 2026Article 50(2) marking for generative AI already on the market, and the ban on nudification apps
2 December 2027Rules for high risk AI systems in Annex III
2 August 2028Rules for high risk AI in products covered by Annex I

Fines

Breaking Article 50 can lead to fines of up to 15 million euros or 3% of total worldwide annual turnover for the previous year, whichever is higher (Article 99(4)). For small and medium sized businesses, including start ups, the cap works the other way round: the fine is capped at whichever of the two is lower (Article 99(6)). National market surveillance authorities enforce the rules, so the first cases will come from the countries where users complain.

Article 50 and your privacy policy

Article 50 sits next to the GDPR, not instead of it. A chat transcript usually contains personal data, so the same conversation that needs an AI disclosure also needs a privacy notice. In practice that means your privacy policy should name the AI service that processes chat messages or calls, say what you use the transcripts for and how long you keep them, and explain whether data leaves the EU. If an outside vendor runs the model, you need a data processing agreement with that vendor.

An AI transparency notice brings the AI specific parts together: which features use AI, what they can and cannot do, how users reach a human, and how AI generated content is labelled. It is the natural place to link from the chat header.

A checklist for a small business

  1. List every place a customer can meet AI on your site, app, phone line and social accounts.
  2. For each one, write down whether you are the provider, the deployer or both.
  3. Add a first message or first sentence that says it is an AI, in plain words.
  4. Add a visible "AI" label to the chat header or voice menu so it stays in view.
  5. Caption AI images and video in marketing where they could pass for real.
  6. Decide who reviews AI written articles before they go out, or label them.
  7. Publish an AI transparency notice and update your privacy policy to name the AI tools that process messages.

You can write both pages with the AI transparency notice generator and the AI content disclaimer generator. For the wider picture of which pages an AI product needs, see legal pages every AI tool needs in 2026.

Frequently asked questions

Yes, if people in the EU can use it and it is not already obvious that it is an AI. The information must reach them at the latest at the first interaction, so put it in the opening message and keep a visible AI label in the chat.

It applies when the output of your AI system is used in the EU. A US business whose chatbot answers visitors in the EU is covered for those conversations, so most businesses show the disclosure to everyone.

No. The disclosure has to be clear, distinguishable and perceivable at the first interaction or exposure, for example a visible or spoken label. Policy pages are a good second place, not the only place.

If you build an AI system or have it built and offer it under your own name, you are the provider. If you use an AI system in your business, you are the deployer. A business that builds its own assistant on a model API and runs it on its site is usually both.

Up to 15 million euros or 3% of worldwide annual turnover, whichever is higher. For small and medium sized businesses, including start ups, the cap is whichever of the two is lower.

The marking duty for providers of generative AI applies from 2 August 2026. Systems already on the market before that date have a grace period until 2 December 2026 under the AI Omnibus.

Only text published to inform the public on matters of public interest, and only when no human reviewed its substance under someone's editorial responsibility. Product descriptions and posts your team reviews are not caught, but labelling AI help is still good for trust.

The bottom line

If a customer can talk to an AI on your site, say so in the first message, keep a visible AI label in the chat, caption realistic AI images, and explain the details in an AI transparency notice. It takes an afternoon and removes one of the easiest fines in the AI Act.

Write your AI transparency notice

Add your website details once and get a notice written for your product, ready to publish next to your chat.

Start with your website
AS
Abd Shanti, FreeTOS.org

Every date, number and quote in this article is checked against the primary source listed below. Research is AI assisted and every claim is reviewed by a person before it is published.

Sources

  1. EU AI Act, Article 50: Transparency obligations for providers and deployers of certain AI systemsThe four duties, the exceptions, and the first interaction and accessibility rule in Article 50(5).
  2. EU AI Act, Article 99: PenaltiesFines up to 15 million euros or 3% of turnover (Art. 99(4)), the lower of the two for SMEs (Art. 99(6)).
  3. EU AI Act, Article 113: Entry into force and applicationApplication dates: 2 February 2025, 2 August 2025 and 2 August 2026.
  4. European Commission: Transparency obligations under Article 50 of the AI Act (FAQ)Provider and deployer definitions, the narrow reading of the obvious exception, visible or audible labels, the human review test and the 2 December 2026 grace period.
  5. European Commission: Opinion on the assessment of the Code of Practice on Transparency of AI generated ContentThe code adequately covers Articles 50(2), 50(4) and 50(5), assessed on 8 July 2026 and confirmed by the AI Board on 9 July 2026.
  6. European Commission: Strong backing for the Code of Practice on Transparency of AI generated Content (31 July 2026)95 provider signatories and 192 deployer signatories, about half small or recent companies.
  7. European Commission: AI Omnibus enters into forceEntry into force on 27 July 2026, high risk dates of 2 December 2027 and 2 August 2028, and the ban on nudification apps.
  8. Regulation (EU) 2026/1744 (Digital Omnibus on AI), Official Journal of the European UnionPublished on 24 July 2026, in force from 27 July 2026; the 2 December 2026 marking deadline for systems already on the market (new Article 111(4) of the AI Act).
Cite this article

“According to FreeTOS.org's guide to EU AI Act Article 50, since 2 August 2026 any chatbot or voice agent that people in the EU can reach must tell them they are talking to an AI at the latest at the first interaction, unless that is obvious. Breaches can be fined up to 15 million euros or 3% of worldwide annual turnover, whichever is higher, and the lower of the two for SMEs. Generative AI placed on the market before 2 August 2026 has until 2 December 2026 to add machine readable marking.”

Source: Shanti, A. (2026). AI Chatbot Disclosure Rules: What EU AI Act Article 50 Requires. FreeTOS.org. https://freetos.org/blog/ai-chatbot-disclosure-eu-ai-act-article-50

This article explains the rules in plain language. It is not legal advice. Last reviewed 27 September 2026.