AI disclosure:What your chatbothas to say

What does Article 50 of the AI Act cover?

Article 50 is headed “Transparency obligations for providers and deployers of certain AI systems”. Under Article 113 of the regulation, it has applied since 2 August 2026. It does not contain a general labelling requirement for every text an AI helped to write. Instead, it describes specific cases, and for each case it sets out clearly who has to act.

Three cases matter most for businesses that deal with customers: direct interaction with an AI system, meaning chat and phone (paragraph 1), machine-readable marking of generated content (paragraph 2) and disclosure of deepfakes (paragraph 4). Paragraph 3 covers emotion recognition and biometric categorisation systems, whose deployers must inform the people exposed to them. Paragraph 5 sets out when and how the information is given: at the latest at the time of the first interaction or exposure, in a clear and distinguishable manner, and in line with accessibility requirements.

What belongs in your chatbot's first message?

Article 50 does not prescribe any fixed wording. What counts is that the notice comes at the latest at the first interaction, is clear and distinguishable, and meets accessibility requirements. A sentence in your terms and conditions or privacy policy does not reach the customer at the moment of the conversation. The notice belongs where the conversation starts.

A greeting that handles three things in one go works well: it says that an AI is answering, what it can help with, and how to reach a human. That turns the obligation into a service, because the customer knows straight away what to expect.

For website chat: “Hello, you are chatting with our AI assistant. I can help with questions about appointments, services and opening hours. If you would rather speak to a member of staff, just type ‘human’.”

For a messenger bot, for example on WhatsApp: “Hello, you are messaging the digital assistant of [business]. This is an AI. I will take down your request and pass it on to our team.”

For a phone assistant: “Hello, you are speaking with the AI assistant of [business]. I will note your request and make sure someone calls you back.” There is no chat window on the phone, so the notice belongs in the spoken greeting, before the first question to the caller.

Less suitable are notices that are easy to miss: a small icon without text, a footnote below the input field, or a name that sounds like a real employee without any mention that it is an AI.

Which AI images, videos and voices must you label?

Under Article 3(60), a deepfake is AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful. Anyone who generates or manipulates such content with an AI system and uses it professionally must disclose that it has been artificially generated or manipulated.

For advertising and social media, this means: an AI photo that looks like a real shot from your workshop, or an AI voice that sounds like a real person, needs a visible or audible notice. For evidently artistic, creative, satirical or fictional works, a notice that does not hamper the display or enjoyment of the work is enough.

AI-generated text published to inform the public on matters of public interest must also be disclosed. This obligation does not apply if the text has undergone human review or editorial control and a person holds editorial responsibility for it.

Who enforces the transparency obligations in Germany?

Under the German AI Market Surveillance and Innovation Promotion Act (KI-MIG), in force since 29 July 2026, the Federal Network Agency (Bundesnetzagentur) is the central point of contact, market surveillance authority and complaints body for the AI Act. In regulated sectors, the existing sector authorities keep their responsibilities, with the Bundesnetzagentur coordinating.

The Bundesnetzagentur has set up an AI Service Desk for companies. There you will find a dedicated page on the transparency obligations, an AI Compliance Compass for initial guidance, FAQs and a contact form. The authority also provides separate guidance for small and medium-sized enterprises and start-ups.

Under Article 99, breaches of Article 50 can be punished with fines of up to 15 million euros or up to 3 percent of worldwide annual turnover. For small and medium-sized enterprises, the lower of the two amounts applies in each case. According to the regulation, all relevant circumstances of the individual case are taken into account when setting the amount, such as the nature, gravity and duration of the infringement.