EU AI Act Transparency Rules for Chatbots: A Practical Checklist
Since 2 August 2026, people must be told when they're talking to an AI. Here's what Article 50 means for customer chatbots, with a practical checklist and wording examples.

In this article
If your business runs a customer-facing chatbot in the EU, the core EU AI Act chatbot rule is simple to state: people must be told they're interacting with an AI system, unless that's already obvious. That requirement sits in Article 50, and its transparency obligations have applied since 2 August 2026.
Applying it across a website widget, WhatsApp, Instagram DMs and a human team that steps in and out of conversations takes more thought. Below: what Article 50(1) asks for, who is responsible, and a checklist you can work through this week.
This article is practical guidance, not legal advice. For your specific situation, consult a qualified lawyer.
Key takeaways
- Tell people they're talking to an AI at the start of every conversation, on every channel, in plain words.
- Don't rely on the "it's obvious" exception for a customer service chatbot.
- Make reaching a human easy, and keep records showing how and where you disclose.
What Article 50 requires from customer-facing chatbots
Article 50(1) says providers must design and develop AI systems that interact directly with people so that those people are informed they're interacting with an AI system. The exception: when it's obvious to a reasonably well-informed, observant person, given the circumstances and context.
Article 50(5) adds two practical details. The information must be given in a clear and distinguishable way, and at the latest at the time of the first interaction. It also has to meet applicable accessibility requirements.
For a customer service chatbot, that translates into: say it up front, say it clearly, and say it in a form everyone can perceive.
Other parts of Article 50 cover marking AI-generated media and disclosing deepfakes. For a text chatbot answering customer questions, 50(1) is the rule that matters most.
Provider or deployer: who is responsible for chatbot disclosure?
The AI Act distinguishes between providers (who develop an AI system and place it on the market or put it into service) and deployers (who use it under their own authority). The Article 50(1) obligation is addressed to providers, which includes the vendor of your chatbot platform.
In practice, the line blurs. You choose the bot's name, avatar, greeting and channels. If you give it a human name and photo, or remove the disclosure from the greeting, you can undermine the vendor's design. And if you build your own chatbot on top of a model API and run it under your brand, you may count as a provider yourself.
The safe working assumption: treat disclosure as your responsibility, and choose a vendor whose product makes it easy to get right.
EU AI Act chatbot checklist
Work through these on every channel where an AI assistant replies.
- Disclose in the first message. Not in the footer, terms or privacy policy: in the conversation itself, before or with the first answer.
- Cover every channel. Web chat, WhatsApp, Instagram DM and any other channel the assistant answers. Disclosure on your website doesn't cover your WhatsApp number.
- Use plain words. "I'm an AI assistant" beats "Powered by advanced conversational technology."
- Disclose in the customer's language. If the assistant answers in Polish, the disclosure should be in Polish too.
- Avoid human impersonation. No human first name plus a stock photo. A friendly bot name is fine if the AI status is clear.
- Make the human route obvious. Tell people how to reach a person, and make the handoff work.
- Signal handovers. When a team member takes over, say so, and say so again if the AI resumes.
- Keep it accessible. Text that screen readers can read, not only an icon or a colored badge.
- Keep records. Screenshots of the disclosure per channel, the date you configured it, and a log of later changes.
- Brief your team. Agents shouldn't pretend to be the bot, or claim a reply was human-written when it wasn't.
For the privacy side, which still applies alongside the AI Act, see our guide to GDPR compliance for AI chatbots.
Chatbot disclosure examples by channel
Good disclosure is short and does three jobs: says it's AI, says what it can help with, and says how to reach a person.
| Channel | Example opening message |
|---|---|
| Website chat | "Hi, I'm the [brand] AI assistant. I can help with orders, delivery and returns. Type 'agent' at any time to reach our team." |
| "Hello, you're chatting with our AI assistant. I can answer questions and book appointments. Want a person instead? Just say so." | |
| Instagram DM | "Thanks for your message. I'm an AI assistant and I'll reply right away; our team can take over whenever you ask." |
| Handover to human | "I'm passing you to [agent name] from our support team, who can see our conversation so far." |
Mistakes that undermine chatbot disclosure requirements
- Disclosure only on one channel, usually the website widget, while WhatsApp replies look like they come from a person.
- A human persona with a first name, profile photo and phrases like "let me check with my colleague" when no colleague exists.
- Burying it in a tooltip, a tiny badge or the privacy policy.
- Silent handovers, where a human takes over and the customer still believes it's the bot, or the reverse.
- No way out. Saying "AI assistant" without offering a route to a person still frustrates customers.
Our guide on writing a chatbot persona shows how to give an assistant personality without implying it's human.
How Conviro supports AI Act transparency
Conviro includes an AI transparency and disclosure pack, built to help businesses meet the EU AI Act's disclosure expectations. It sits alongside defaults that were already part of the platform: human handoff is always possible, conversations land in a shared inbox with the full transcript, and the AI assistants answer in the customer's language, so the disclosure can too.
On the data side, Conviro is GDPR compliant, hosts data in the EU in ISO 27001-certified data centers, offers a DPA under GDPR Article 28, and doesn't use customer data to train models. The security page has the details.
Frequently Asked Questions
Does the EU AI Act require chatbots to say they're AI?
Yes. Article 50(1) requires AI systems that interact directly with people to be designed so those people are informed they're interacting with AI, unless it's obvious from the context. For customer service chatbots, the practical answer is to disclose clearly in the first message on every channel.
When do the AI Act transparency rules apply?
The transparency obligations in Article 50 apply from 2 August 2026. If your chatbot launched before that date, it still needs to meet them now. Check your existing greetings, channels and handover messages rather than assuming an older setup already complies.
Is a customer service chatbot a high-risk AI system?
Usually not. Typical customer service chatbots aren't among the high-risk uses listed in the AI Act, though a system used for decisions such as credit scoring or hiring could be. Even when a chatbot isn't high-risk, the Article 50 transparency rules still apply.
What should a chatbot disclosure message say?
Keep it to one or two sentences: that the customer is talking to an AI assistant, what it can help with, and how to reach a person. Write it in the customer's language, show it at the start of the conversation, and avoid human names or photos that suggest otherwise.
Next steps
Open each channel where your assistant replies and read the first message a new customer sees. If it doesn't clearly say "AI," fix that first, then work through the rest of the checklist. Start free with Conviro to set up an assistant with disclosure and human handoff built in.
Published by
Conviro
Sharing insights on AI-powered customer support, chatbot automation, and SaaS growth strategies.
Put the playbook into practice
Turn your support knowledge into faster answers
Explore Conviro’s AI support workspace, then start free when the workflow fits your team.


