Straight answers6 min read

The EU AI Act and customer service chatbots: what applies

Short answer

Since 2 August 2026, Article 50 of the EU AI Act requires that anyone chatting with an AI is told so, at the latest at first contact, unless it is obvious. A customer service chatbot is normally not a high-risk system, so the transparency duty is what applies. In Supportifier you write the welcome message yourself, AI answers are labelled Generated by AI by default, and the visitor can ask for a person.

Rickard CollanderFounder, Supportifier
Published
Updated

The short answer: Yes, the customer must be told they are talking to an AI. The requirement is in Article 50 of the EU AI Act and has applied since 2 August 2026. The information must come at the latest at first contact and be clear. A customer service chatbot is normally not a high-risk system, so for most support teams this is the part of the Act that shows up in daily work. This page is not legal advice; check your setup with a lawyer.

What Article 50 requires

The AI Act, Regulation (EU) 2024/1689, is risk-based: some uses are prohibited, high-risk systems carry extensive requirements and a wider group of systems has transparency duties. Chatbots belong to the last group.

Under Article 50(1), an AI system intended to interact directly with people must be designed so that the person is informed that it is an AI system. The exception is when this is obvious to a person who is reasonably well-informed, observant and circumspect. The European Commission's questions and answers on Article 50 say that exception is to be read restrictively. Do not count on the customer working it out.

Article 50(5) says how: the information must be given in a clear and distinguishable manner, at the latest at the time of the first interaction, and meet the applicable accessibility requirements. In a chat that means a text in the chat window itself when the conversation starts, not a line in the privacy policy.

What applies when

WhatFrom
The customer must be informed that they are interacting with an AI (Article 50(1))2 August 2026
Machine-readable marking of AI-generated content, for systems on the market before 2 August 2026 (Article 50(2))2 December 2026
The obligations for high-risk systemsPostponed to December 2027 and August 2028

The changes in the 2026 Digital Omnibus did not move the duty to say that it is an AI. The high-risk dates come from the law firm Gibson Dunn's review of the agreement; the period to 2 December is in the Commission's questions and answers.

According to the Commission, breaches of the transparency duties can lead to fines of up to 15 million euros or 3 percent of worldwide annual turnover. Enforcement is mainly handled by national market surveillance authorities.

Who is responsible: the vendor or you?

The Act separates the provider, which develops the AI system and places it on the market, from the deployer, which uses it in its business. Article 50(1) is addressed to the provider: the system must be built so that the person is informed. You, who put the chat on your website, decide in practice what the customer meets: the name of the bot, the welcome text and whether answers are labelled. A human name and a portrait on the bot make it harder to argue that the customer was informed.

So ask your vendor how the chat says that it is an AI, and check for yourself how it looks on your site, on a phone and with a screen reader.

How it looks in Supportifier

  • The welcome message in the AI chat is written by you, along with the chat's heading. That is the place for "You are chatting with our AI assistant".
  • The label Generated by AI is shown on answers in the chat and in the AI contact form. It is on by default and is controlled in the settings for the widget's appearance.
  • A way to a person. After every AI answer the visitor can send a message that becomes a ticket, and when live chat is staffed choose to talk to a person. That is not an explicit requirement in Article 50, but it is what makes the transparency worth something to the customer.
  • In the inbox the default is a suggestion that an agent reads, edits and sends. Then it is the agent who answers the customer. Automatic replies are switched on per ticket category and can first run in shadow mode, where you see what would have been sent.

The label is on by default, but the welcome text is yours to write. Leave the label on and write the welcome message before the chat goes live.

Checklist for your chat

  1. Say in the chat's first message that it is an AI, what it can help with and how the customer reaches a person.
  2. Give the bot a name that cannot be mistaken for a member of staff.
  3. Leave the labelling of AI answers in the chat and in the form switched on.
  4. Decide with your lawyer how automatic email replies should be labelled, if you switch them on.
  5. Test the flow with a screen reader and on a phone, since the information must also be accessible.
  6. Write down what you did and when, so you can show it if a customer or an authority asks.

Questions buyers ask

Does a chatbot have to say it is an AI? Yes. Article 50 requires that the person is informed at the latest at first contact, unless it is obvious, and the Commission writes that the exception is to be read restrictively. A clear text when the chat opens is the simplest way.

Is a customer service chatbot a high-risk system? Normally not. The high-risk list covers areas such as employment, credit scoring and access to essential services. If the AI is used to make such decisions about a customer, that is a different assessment, and a lawyer should look at it.

Does the duty cover AI drafts that an agent sends? Article 50(1) is about systems that interact directly with people. A draft that an agent reads, edits and sends in their own name is an answer from the agent. If the AI sends the reply itself the situation is different; check it with your lawyer.

Is a line in the terms enough? No. The information must be given clearly at the latest at the first interaction, which means where the conversation takes place.

What does the GDPR have to do with it? The GDPR applies to the personal data in the conversation whoever answers, and is a separate set of rules. We go through it in the article on GDPR and AI in customer service and in the answer on storage within the EU/EEA.

Read more in the answer on AI chat with handoff to a person and on the security page.

Sources

Sources checked on 21 November 2026.

This page is also available in Swedish: AI-förordningen och chattbotar i kundservice: vad som gäller

Next step

Start from your everyday work.

Tell us which question takes time today. We go through what a first step could look like.

A clear first step beats a big promise.

See how the knowledge work, the review and the first channel fit together.

About the knowledge analysis
From the same question
to a better answer.
Book a walkthrough