You Have a Customer Chatbot. Here's Exactly What the EU AI Act Requires.

The Most Common AI System in Business

Customer chatbots are the most widespread AI deployment in EU SMEs. They're on e-commerce sites, SaaS platforms, support portals, and local business websites. They handle enquiries, book appointments, answer FAQs, and escalate complaints.

They are also specifically addressed by the EU AI Act — and from August 2, 2026, your chatbot must comply with Article 52's transparency obligations if it's accessible to EU residents.

What Article 52 Actually Says

Article 52(1) of the EU AI Act states:

"Natural persons who interact with AI systems intended to interact with natural persons shall be informed in a clear and timely manner that they are interacting with an AI system, unless this is obvious from the circumstances and the context of use."

Plain English: customers must know they're talking to AI. The disclosure must be clear (not buried in fine print) and timely (before or at the start of the interaction, not at the end).^[1]

The "Obvious" Exception — and Its Limits

The regulation includes an exception: if it's obvious to a "reasonably well-informed user" that they're interacting with AI, no disclosure is required.

What qualifies as "obvious"? A clearly labelled robot avatar. A chatbot named "AI Assistant" in prominent lettering. An automated phone menu that begins "You're speaking with our automated assistant."

What doesn't qualify: a chatbot with a human name ("Hi, I'm Sophie!") that engages in conversational language without any indication it's automated. Even if you know it's AI, the exception only applies if it's obvious to the average customer encountering it for the first time.

When in doubt, disclose. The penalty for unnecessary disclosure is zero. The risk for missing a required disclosure is not.

The Practical Fix

For the vast majority of chatbots, compliance is a five-minute change. Options include:

Set the chatbot's first message to: "Hi! I'm an AI assistant. How can I help you today?"
Add a persistent label in the chat widget header: "AI-powered support"
Include a disclosure line in the chatbot's introductory prompt

Any of these approaches satisfies the "clear and timely" requirement. You don't need to explain how the AI works or what model it uses — just that it is AI.

AI-Generated Content: A Different Obligation

Article 52(3) addresses a different but related requirement: AI-generated images, audio, and video must be labelled as AI-generated. This applies to:

AI-generated product images or marketing visuals
Synthetic voice recordings (customer service IVR using AI voice)
AI-generated video content published publicly
Deepfake-style content of any kind

The disclosure must be machine-readable where technically feasible, and human-readable in the interface. For marketing content, this typically means a small label or footer: "Image generated with AI" or "Voice: AI-synthesised."^[2]

When Do These Requirements Apply?

For existing chatbots and AI systems deployed before August 2, 2026: the deadline is August 2, 2026. You have until then to update your systems.

For chatbots deployed after August 2, 2026: the rules apply from day one of deployment.

There is one important note about emotion recognition AI: using AI to detect or infer emotions of individuals in workplace and educational settings has been prohibited since February 2, 2025. This is not a chatbot transparency issue — it's a prohibited practice. If your customer service AI analyses customer sentiment or emotion to route calls or adjust pricing, check carefully whether it crosses this line.

Is This All You Need to Do?

For a standard customer service chatbot, adding an AI disclosure is usually all that's required under the EU AI Act. Customer-facing chatbots are classified as limited risk — they carry real obligations, but those obligations are primarily about transparency, not the full compliance apparatus required for high-risk AI.

The exception: if your chatbot makes decisions about people (eligibility for services, pricing, access), the risk classification may be higher, and the obligations more demanding. Use our free audit to confirm your chatbot's risk category and get a personalised compliance checklist.