AI disclosure to visitors

Visitors should always know when they're talking to an AI rather than a person. It's fair, it sets the right expectations for the answers, and in the European Union it's the law. The ASDK widget tells every visitor, on every site, and you can't switch that off. This page explains the rule and what's left for you to do.

What the EU AI Act asks for

Article 50 of the EU Artificial Intelligence Act (Regulation (EU) 2024/1689) sets transparency rules for AI systems. Its first paragraph covers systems that interact directly with people, such as chat assistants: they must be designed so that the people using them are informed that they are interacting with an AI system, unless that is obvious from the context.

Paragraph 5 adds how: the information must be given in a clear and distinguishable manner, at the latest at the time of the first interaction, and it must meet accessibility requirements.

Article 50 applies from 2 August 2026, the Act's general date of application under Article 113. The text of Article 50 and Article 113 is on the European Commission's AI Act Service Desk, and the full regulation is on EUR-Lex (all checked October 2026).

The duty to design the system this way sits mainly with the provider, which for the widget is ASDK. But you decide how the widget appears on your site, so it's worth knowing what keeps it compliant.

How the widget discloses it

• A disclosure line under the title, always visible while the panel is open, before the visitor types anything. The default is "AI assistant · answers can be wrong, check important details".
• In the visitor's language. The default line comes in the widget's eight interface languages.
• For screen readers too. The panel is described by the disclosure line, so assistive technology announces it when the panel opens.
• It can't be removed. You can reword it on the Agents page, but an empty field restores the default.

Answers also cite the pages they came from, which helps visitors check what the AI said.

What you can change

In the Messages section of the Agents page, the AI disclosure field takes up to 160 characters. Reword it to match your tone, but keep it unmistakable:

Your own text is shown to every visitor as written. If your visitors speak several languages, keeping the default means each sees it in their own language. See Widget texts and greeting.

Good practice beyond the minimum

• Say it in the greeting too, for example "Hi! I'm the AI assistant for Acme." Visitors read the greeting first.
• Don't give the agent a human persona. A human name and photo invite visitors to believe they're talking to a person.
• Offer a person. Lead capture lets visitors leave their details for a human reply. See Lead capture.
• Mention the chat in your privacy notice: that an AI assistant answers questions, what's stored and for how long. See Data we store and Data retention.

Outside the EU

Other places have their own rules on automated chat, and they differ in scope. A clear AI disclosure is a sensible default wherever your visitors are, and the widget shows it everywhere. Our article on AI chat and privacy discusses what to tell visitors in more detail.

What ASDK does on its side

• The disclosure is part of the widget for every customer and every plan.
• ASDK doesn't make decisions about visitors with legal or similarly significant effects; agents answer questions and pass requests to your team.
• AI model providers process visitors' questions under terms that don't allow training on them. See No training on your data.

Related: Subprocessors, the glossary entry on AI agents.