Guides · Updated 2026-10-22

Intercom and Fin: what to check for the AI Act's chatbot notice

Intercom’s Fin answers support questions on its own before, or instead of, a person. On an EU-facing site that makes the Messenger an AI system that interacts with people, so Article 50(1) of the EU AI Act applies from the moment a visitor starts a conversation.

The split of duties

Intercom, as the provider, must make it possible to tell people they’re talking to an AI. You, as the deployer, configure the Messenger on your site and are the one a visitor or authority will ask. The practical test is simple: open your own site in a private window, start a chat, and see whether an ordinary person would know they’re talking to an AI before they type.

What to verify

1. Fin’s identity in the Messenger. Does it introduce itself as an AI, or does it use a name and avatar that could be a person? An introduction like “Fin, AI agent” works; “Anna from Support” with a stock photo does not.
2. The first message. Fin’s opening line should make the AI nature clear in the visitor’s language.
3. Localisation. Intercom supports multiple Messenger languages; confirm the AI wording is translated, not only the buttons.
4. Hand-over to humans. Fine, as long as the notice was shown at the start.
5. Email replies. If Fin also answers inbound email, the same duty applies there; the notice belongs in the email itself.

Keep on file

Dated screenshots of the Messenger settings and of a real first message in each language; Intercom’s published position on EU AI Act transparency, if any; your last review date.

Intercom and Fin are trademarks of their owner. This guide is independent, describes what to check in your own setup, and is not legal advice.

General information, not legal advice. Check the Regulation text and the Commission's guidelines, or ask a lawyer, for your case.