EU AI Act Transparency Rules Begin: What AI Chatbot Providers Must Change

The European Commission has begun enforcing new EU AI Act transparency requirements that affect AI systems which interact directly with people, including chatbots. From 2 August 2026, providers of covered systems must ensure users are told they are interacting with AI, generally at the first interaction. For companies using conversational AI in customer service, sales or marketing, the change makes clear user-facing disclosure a product and implementation requirement, not simply a matter of good practice.

The obligation is grounded in Article 50 of the AI Act and is not a blanket rule that every automated interaction must use identical wording. Its scope includes exceptions, notably where it is obvious to a reasonably well-informed person that they are dealing with an AI system. Still, businesses should not assume that a chatbot's name, interface or use case will automatically satisfy that exception. The European Commission's enforcement announcement also links the start of enforcement to wider transparency obligations for AI-generated and manipulated content.

What Article 50 requires

Article 50 requires providers of AI systems intended to interact directly with natural persons to design those systems so that people are informed they are interacting with an AI system. The information must be provided no later than the first interaction, unless the AI nature of the interaction is obvious from the circumstances and context to a reasonably well-informed person.

For a customer-facing chatbot, the practical point is straightforward: a visitor should receive a clear, understandable notice before or when the conversation starts. This can be incorporated into the chat interface, opening message or another prominent part of the interaction. The law's focus is on informing the person, rather than prescribing one standard sentence or visual design.

The Commission's materials also cover separate transparency duties for AI-generated or AI-manipulated content. That content must be marked clearly and made detectable, including through machine-readable marks. In addition, deployers of certain AI systems used for emotion recognition or biometric categorisation must tell exposed individuals that the system is operating.

The chatbot rule has an important exception

The exception for interactions that are obviously AI-driven is significant, but narrow in practical terms. It depends on what would be obvious to a reasonably well-informed person in the specific context. A clearly labelled experimental bot may present a different case from an assistant embedded in a support widget that resembles a normal human service channel.

The verified material does not define a universal checklist for deciding whether an interaction is obvious. That means providers should examine the customer journey, presentation and wording of each system rather than rely on a generic assumption about chatbots. The Commission has published accompanying guidance and points to a Code of Practice on AI-generated content to support implementation.

Transition period for existing generative AI systems

The Commission's Quick Facts material identifies a transition period for generative AI systems placed on the market before 2 August 2026. Their labelling obligations have a grace period until December 2026. This date matters for existing tools, but it should not be read as a general delay for all transparency duties or every type of AI system.

For new and existing deployments alike, the first task is to identify whether the organisation is the provider of the system, a deployer using another provider's product, or both. The Article 50 chatbot obligation described here is directed at providers. Contractual responsibilities, product configuration and communication with customers may nevertheless make the issue relevant to companies that deploy third-party chat tools.

Turning the rules into practical implementation work

A useful implementation approach begins with the systems customers and other individuals can actually encounter. Support chatbots are an obvious priority, but teams should also review lead qualification assistants, website help widgets, conversational booking tools and AI features inside customer portals.

Three practical checks can bring structure to that review:

• Map direct interactions: identify every system designed to converse or otherwise interact directly with natural persons.
• Review the first touchpoint: check whether a person receives a clear notice that the system is AI when the interaction begins.
• Separate content obligations: inventory AI-generated or manipulated content and determine how clear and machine-readable marking will be handled.

This is not merely a legal text exercise. Clear disclosure can set appropriate expectations about a tool's role and limitations before a customer relies on it. It can also prevent a conversational interface from being presented in a way that leaves users uncertain whether they are speaking with a person.

For businesses, the operational work may involve product owners, web teams and the supplier of a chatbot platform. If a third-party system controls the opening message or interface, the customer-facing business may need to determine what can be configured and what documentation the supplier provides. Where AI-generated material is part of a marketing or service workflow, the content-production process may also need a reliable way to preserve required marking.

The Commission's enforcement milestone makes this a timely point to audit live AI touchpoints and planned launches. The research supplied here confirms the core obligations and transition period, but does not provide a complete legal analysis for every deployment. Organisations with uncertain or high-impact use cases should seek advice tailored to their circumstances.

For businesses rolling out customer-facing AI, disclosure is only one part of making a tool work reliably in real workflows. Scalevise can help assess use cases, identify implementation gaps and turn policy requirements into practical product decisions through AI implementation and adoption consultancy. A focused review can reduce rework, clarify which customer journeys need changes and help teams deploy AI with clearer user communication. Request an AI implementation consultation.

Frequently Asked Questions

When did the EU AI Act chatbot transparency rules begin to be enforced?

The European Commission announced that enforcement of the relevant AI Act rules began on 2 August 2026, coordinated with national authorities.

Do all chatbots have to disclose that they are AI?

Providers of AI systems intended to interact directly with natural persons must inform users that they are interacting with AI, unless this is obvious to a reasonably well-informed person from the circumstances and context.

When must a chatbot tell a user that it is AI?

The information must be provided no later than the first interaction, subject to the exception where the AI nature of the interaction is obvious.

Is there a grace period for generative AI labelling?

Yes. For generative AI systems placed on the market before 2 August 2026, the Commission states that labelling obligations have a transition period until December 2026.

What other transparency duties are covered by the new rules?

The Commission says AI-generated or manipulated content must be clearly marked and detectable, including with machine-readable marks. Certain emotion recognition and biometric categorisation deployments also require notice to exposed individuals.

Conclusion

The start of enforcement gives Article 50's transparency requirements immediate practical relevance for AI providers and the businesses building customer-facing AI experiences. A clear disclosure at first interaction, careful treatment of AI-generated content and attention to the applicable exceptions and transition period can help organisations align their tools with the EU's new transparency framework.