From 2 August 2026, the EU AI Act's transparency rules apply to any organisation using AI to interact with customers. The rule is simple: if a customer is talking to AI, they need to be told clearly, from the very start of the conversation.
This applies to everyone using AI on chat, messaging, or voice, and it's easy to get right. Good news too: Talkative's AI is classed as limited risk under the Act, so for most contact centres this is about tightening a few settings, not rebuilding your setup.
⚠️ The key principle: the disclosure has to be visible in the conversation itself. A note buried in your terms and conditions or privacy policy is not enough on its own.
Below is a short checklist of what to do. Most items are copy-and-paste ready - just swap in your own names.
1. Give your AI a name that shows it's AI
If your chatbot has a name, that name needs to make clear it's AI, because it appears on every message the customer sees.
Add "AI" to the name. "Assistant" or "Virtual Assistant" on their own are not enough - the word "AI" needs to be there.
Examples:
Josh → Josh AI Assistant
Talkative Chatbot → [Your Company] AI Assistant
Why it matters: the name is on screen for the whole conversation, so it's the simplest way to keep the disclosure visible throughout.
2. Say it in your welcome message
Make the first thing your AI says a clear introduction that it's AI.
Example welcome message:
Hi, I'm Josh, [Your Company]'s AI assistant. I can help with your questions, and I'll connect you to a member of the team whenever you need one.
Why it matters: people often reach out when something is already urgent or frustrating. Setting expectations up front builds trust - and it sets up a smooth handover to a human when one's needed.
3. Add an AI-disclosure line to your AI prompts
In your AI prompt/instructions, add a line telling the AI to confirm it's AI if a customer ever asks.
Example prompt line:
If a customer asks whether they are speaking to a human or to AI, always confirm clearly that you are an AI assistant.
Why it matters: this covers you if a customer asks directly mid-conversation. The AI should never imply it's human.
4. Do the same for Voice AI
The rule applies to voice as well as chat. State it in the spoken intro at the start of the call, or in your pre-call disclaimer if you use one.
Example voice intro:
Hi, you're through to [Your Company]'s AI assistant. I can help with your call, and I'll pass you to a colleague if you need one.
Why it matters: callers can't see a name on screen, so the spoken introduction is the only place the disclosure lands.
5. Review your website policies (good practice)
Check your privacy policy and any relevant website terms are up to date on how you use AI and handle conversation data.
Why it matters: this is sensible governance and worth doing - but note it does not replace the steps above. The Act requires the disclosure to be clear inside the conversation, so your policy pages are a complement to that, not a substitute.
📚 Further reading
For the full picture, how Talkative is preparing, why our AI is classed as limited risk, and a broader compliance checklist… read our guide:
If you'd like help reviewing your own setup ahead of the deadline, get in touch with your Talkative CSM.