Chatwoot and the EU AI Act: does your chat widget tell people it's AI?
Chatwoot is open-source — self-hosted or Chatwoot Cloud — which means its Captain AI Agent shows up on sites run by developers and small teams who deployed it themselves, often without anyone else reviewing the setup. Captain works as named, configurable "Assistants" that can be assigned per inbox (a support Assistant on the website widget, a sales Assistant on WhatsApp, and so on) — flexible, and one more setting screen that can quietly ship without a visitor-facing AI disclosure.
The rule itself is short. Article 50(1) of the EU AI Act requires AI systems that interact directly with people to be designed so users know they're dealing with AI — no later than the first interaction, in a clear and distinguishable way. The exception for cases where it's "obvious" is narrow: a natural-language customer-service bot doesn't qualify just because it has a robot icon. A line in your terms of service doesn't satisfy it either. And it's easy to enforce, because a regulator can simply open your site and start a chat.
Where the disclosure lives in Chatwoot
In a typical Chatwoot setup, the surfaces that can carry (or fail to carry) the AI disclosure are:
- The website widget launcher and greeting for whichever inbox has a Captain Assistant assigned
- The Assistant's configured name and behavior — set per inbox, so different channels can present differently
- Captain's first generated reply, drawn from your knowledge base and past conversations
- The handoff point where a conversation moves from Captain to a human agent in the same inbox
Your Chatwoot disclosure checklist
- Open the widget for each inbox with a Captain Assistant and start a conversation: does the first reply identify itself as AI?
- Check the Assistant's configured name — Captain's setup asks for an internal "Assistant Name," so confirm separately what visitors actually see and that it isn't a bare human-sounding label
- Add an explicit AI statement to the Assistant's greeting or system instructions, since a self-hosted default won't necessarily include one
- If you run multiple Assistants across inboxes (support, sales, WhatsApp), check each configuration independently — per-inbox flexibility means per-inbox drift
- Verify the human-agent handoff is visible to the visitor as a participant change, not just an internal inbox reassignment
- Screenshot each inbox's first-interaction state and keep a dated export of your Assistant configurations
Watch out for
Because Chatwoot is self-hosted or lightly-managed cloud software aimed at technical teams, Captain often gets configured once by a developer optimizing for accuracy and tone, with no compliance review step in the deployment process at all. Open-source flexibility is the same property that lets a disclosure gap ship silently and stay unnoticed indefinitely.
Common questions
We self-host Chatwoot — does that change our obligations?
No — Article 50(1) attaches to the interaction with EU visitors, not to who hosts the software or whether you paid for it. Self-hosting removes Chatwoot the company from the equation, but not you.
We run a different Captain Assistant per inbox — do we need to disclose in all of them?
Yes, independently. Each Assistant is a separate configuration and a separate first-interaction moment; fixing your website widget's Assistant does nothing for the one answering on WhatsApp.
Check it in one scan.
DisclosureProof opens your site the way a regulator would, triggers your Chatwoot widget, and records whether the AI disclosure actually appears — with timestamped evidence either way.
Scan your site free