Do AI chatbots have to say they're AI? A plain-English guide to the disclosure rules
By The Relay team · Published · 5 min read
The short answer
Often, yes. The EU AI Act requires AI systems that talk to people to tell them it's an AI, unless that is obvious, and has applied since 2 August 2026. California makes it unlawful to use a bot to mislead people about its artificial identity in order to sell to them, unless you disclose it's a bot. Utah requires disclosure when asked, and up front in some regulated settings. Labelling your support bot as AI from the first message is the simple, safe practice. This is general information, not legal advice.
The short version by place
| Where | What it says | Applies |
|---|---|---|
| European Union (AI Act, Article 50) | AI systems meant to interact directly with people must inform them they are interacting with an AI system, unless that is obvious to a reasonably well-informed, observant person in the context. The information must be given clearly at the latest at first interaction | Since 2 August 2026 |
| California (Business and Professions Code, section 17941) | Unlawful to use a bot to communicate online with a person in California, with intent to mislead them about its artificial identity, to knowingly deceive them in order to encourage a sale or purchase, or to influence a vote. Not liable if you disclose that it's a bot. The disclosure must be clear, conspicuous and reasonably designed to inform | In force |
| Utah (Artificial Intelligence Policy Act, as amended in 2025) | Disclose that a person is interacting with generative AI when they clearly and unambiguously ask; stricter, proactive disclosure applies to regulated occupations and certain high-risk interactions | In force |
European Union
Article 50(1) of the AI Act says providers must ensure that AI systems intended to interact directly with natural persons are designed so those persons are informed they are interacting with an AI system, unless that is obvious to a reasonably well-informed, observant and circumspect person given the circumstances. Article 50(5) says the information must be clear and distinguishable and given at the latest at the first interaction. The European Commission's FAQ states that Article 50 applies from 2 August 2026 (so it applies now) and that the "obvious" exception is read narrowly.
The FAQ places the duty to design chatbots so that people are informed on the provider of the AI system; a business that deploys someone else's chatbot can have other duties. Which role you hold in a given setup is a legal question.
California
Section 17941 isn't a blanket rule that all bots must announce themselves. It makes it unlawful to use a bot to deceive someone about its artificial identity in order to incentivise a purchase or sale, or to influence a vote, and says you aren't liable if you disclose that it's a bot. The disclosure has to be clear, conspicuous and reasonably designed to inform. A support chat on a product site that offers to sell you something is the kind of setting the law was written about, so disclosing is the safe practice.
Utah
Utah's AI Policy Act requires a business that uses generative AI to talk to consumers to disclose that fact if the consumer clearly and unambiguously asks. The 2025 amendments focus proactive disclosure on regulated occupations and high-risk interactions, such as those involving health, financial or biometric data or advice in those areas. Summaries by law firms and the Future of Privacy Forum are listed below.
What good disclosure looks like
- Say it before the first answer, not after a customer asks.
- Keep saying it. Label each bot message, not just the window.
- Make the label truthful. If answers come from keyword search rather than a generative model, say "automated assistant", not "AI".
- Answer honestly if asked. A bot should never claim to be a person.
- Make the human path visible, and stop the label once a person takes over, so the label always matches who is talking.
Relay's widget does this: a notice under the header ("You're chatting with an AI assistant, not a person. A human is one tap away."), an "AI assistant" label on every bot message and in the header, "automated assistant" wording when answers come from keyword search only, and no label once a human takes over. That is a design practice, not a statement that your use of it meets any law. See how handoff works.
Questions
Common questions
- Do AI chatbots have to disclose that they are AI?
- In the EU, AI systems that interact directly with people must inform them, unless it is obvious, since 2 August 2026. In California, using a bot to deceive people about its artificial identity to make a sale is unlawful unless you disclose. In Utah, disclosure is required when a consumer clearly asks. Rules vary by place, so check yours.
- When does the EU AI Act chatbot rule apply?
- The European Commission's FAQ says Article 50 of the AI Act applies from 2 August 2026.
- Is it safer to always label a support bot as AI?
- Yes, as a practice. A clear label from the first message, on each bot message, avoids the main risk in the rules above, which is a customer not knowing what they are talking to.
- Does Relay's bot say it is an AI?
- Yes. The widget shows a notice that the customer is chatting with an AI assistant and labels each bot message "AI assistant". If answers come from keyword search only, it says "automated assistant". The labels go away when a human takes over.
Sources and how we wrote this
Relay is made by the team that wrote this post. Figures for other tools come from their own public pages, listed under Sources, and are list prices in US dollars that can change. If something here is out of date, tell us on the feedback board and we'll correct it.