Regulation & AI

When the Chatbot Has to Identify Itself: AI Transparency in Customer Service from August 2026

· 2 min read

From 2 August 2026 the transparency obligations under Article 50 of the EU AI Act become enforceable. What the disclosure duty for chatbots and the labelling of AI-generated content mean in practice for customer service.

Automated first responses, AI-assisted ticket triage, generated reply suggestions: in customer service, AI is already routine. From 2 August 2026 Article 50 of the EU AI Act becomes enforceable, bringing binding transparency obligations for exactly these systems. These duties are distinct from the rules for providers of general-purpose AI models (GPAI): the point here is not the model, but the interaction with people.

What Article 50 requires

At its core sits a disclosure duty. Providers of AI systems intended for direct interaction with people (such as chatbots) must ensure that the person concerned is aware they are communicating with an AI system. The notice must be given at the latest at the time of the first interaction, unless the use of AI is obvious from the circumstances. On top of that, AI-generated or manipulated content must be marked as such in a machine-readable way, and content of public interest that could appear authentic has to be labelled separately.

What “informed” means in practice

A line in the legal notice is not enough. The disclosure has to be clear, given before or at the start of the conversation, and understandable to the user. In practice that means more than a banner:

  • Visibility: The AI notice sits at the start of the dialogue, not in a collapsible footnote.
  • Language: The notice appears in the language the person uses to make contact, not only in English.
  • Handover: When the conversation moves between AI and a human agent, it should stay clear who is currently speaking.

More than a notice banner

The harder side is demonstrability. Anyone who wants to show, when the regulator asks, that the notice was actually and consistently displayed needs a record of the configuration and of the notices served. In multi-tenant environments there is an added layer: each tenant runs its own channels, languages and escalation paths, so the transparency logic has to work per tenant, not just globally. A customer service hub like Datargo CRM, which brings live chat, tickets and contacts together in a multi-tenant way, offers the natural place to anchor disclosure, language mapping and human handover consistently.

Article 50 does not ban AI in customer service. It only requires that no one is left in the dark about who, or what, they are talking to.

Back to the blog