Do I have to tell customers when AI wrote something?
If a chatbot or assistant talks to people, they must be told it is AI — from 2 August 2026.
Drafted with AI help and checked against the official EU text before publishing.
If a chatbot or assistant talks to people, they must be told it is AI — from 2 August 2026. If you publish AI-made images, audio, video or public-interest text, that content needs a clear mark too.
You don't need the chatbot notice where it is already obvious that people are talking to an AI. For written text, a label is only needed when AI-generated text is published to inform the public on matters of public interest, and not when a person has reviewed it and takes editorial responsibility for it.
Telling staff internally that a draft was AI-assisted is good practice. The legal mark is about people who interact with the system or see published content.
This is general information, not legal advice.
What the law actually says
"Providers of AI systems, including general-purpose AI systems, generating synthetic audio, image, video or text content, shall ensure that the outputs of the AI system are marked in a machine-readable format and detectable as artificially generated or manipulated."— EU AI Act, Article 50(2) (Regulation (EU) 2024/1689); applies from 2 August 2026 · eur-lex.europa.eu
In plain words
- Article 50(1) covers tools that talk to people: the person should know they are interacting with AI, unless that is already obvious.
- Article 50(2) is mainly on the provider of the generating tool: outputs should be marked in a machine-readable way.
- If you publish a deep fake, Article 50(4) says you (the deployer) must disclose that it was generated or manipulated. The same article has a narrower rule for public-interest text.
- Telling staff internally that a draft was AI-assisted is good practice. The legal mark is about people who interact with the system or see published content.
- Will look for signs of AI use in the folders you choose, and write them down for you.
- Will write a plain-language readiness report. No evidence reads "Not assessed", never a green tick.
- Will run as a free Windows program on your own computer and send nothing from it.
- Coming with Guard: a check every month, kept so you can see your history and what changed since the last one.
- Coming with Guard: a printable readiness pack built from your check results — AI use found, training notes, disclosure checklist.
Did this answer your question?
Readiness guidance, not legal advice. The Free Toolkit sends nothing from your computer.