Do you have to disclose that your chatbot is AI? Certain AI systems are subject to transparency duties. It depends on the system and the context. A customer-facing chatbot is a typical signal that a transparency review is warranted. It is not automatic. Using ChatGPT, Copilot or Gemini at work does not classify the company by itself.
This page explains what Article 50 of Regulation (EU) 2024/1689 covers, what not to take as given, and where to start. The official source is Article 50: EUR-Lex (ELI). Definitions: AI Act glossary.
What Article 50 covers (and what it does not)
Article 50 sets transparency duties for certain AI systems. It is not a generic duty to “always disclose that there is AI”. Classifying a concrete case requires review.
In practice, companies usually ask two things:
- Whether people must be informed that they are interacting with a chatbot or an AI system.
- Whether content generated or manipulated with AI (text, images, audio or video) must be identified.
Those questions are the starting point of a transparency review. They are not a confirmed duty. They are not a definitive legal classification.
What it is not (and why that matters)
- Having a chatbot does not, by itself, confirm a transparency duty. It opens a review. The user flow, who interacts and the company’s role matter.
- Using ChatGPT, Copilot or Gemini does not automatically make every company an AI provider. Determine the potential role first: provider, deployer or other.
- Generating content with AI does not, by itself, open a concrete transparency duty. It is a signal. It requires review in your company’s context.
- Article 50 is not a certificate, a seal, or an automatic readiness score.
The Assessment identifies the use; it does not declare whether the AI Act applies to your case.
Calendar already published: the Omnibus did not postpone Article 50
The AI legal risk assessment guide and the 2026 AI Act guide already state the timeline we use:
- 2 August 2026: broader application, including Article 50 transparency.
- 2 December 2026: not a generic “new prohibitions package” for every company. It is a narrow change. Those same guides mention, for generative systems already on the market, the technical marking duty in Article 50(2).
The Digital Omnibus moved Annex III high-risk duties to December 2027. It did not affect Article 50 transparency duties, which still apply from August 2026. If your company runs a customer-facing chatbot, that transparency track has not been delayed. Whether it applies to your concrete case requires review.
Chatbots, generated content and internal use
A customer-service chatbot and an internal tool for drafting reports do not have the same profile. The first often opens a transparency review. The second may not. In both cases the GDPR and the provider contract can still be relevant.
The AI Act applies alongside other relevant EU rules, including data protection. Compliance with the AI Act does not replace compliance with the GDPR.
If the bot comes from a vendor, also look at the SaaS contract: what it says about notices, marking, logs and liability. That is not, by itself, “the Article 50 obligation”.
Where to start
Legal Stones may recommend legal review when the information is insufficient or the case is complex. The first step is usually to see real use.
- Snapshot of use: five questions (ChatGPT, client data, policy, inventory, personal tools).
- If the signal is critical or moderate, the AI Legal Assessment: inventory, needs and a documented plan. Remote. It is not a definitive legal classification.
- If the provider contract is already on the table, contract review.
- Or write to us and we will point you to the next step.
Frequently asked questions
Do I have to disclose that my chatbot is AI?
Certain AI systems are subject to transparency duties. It depends on the system and the context. A customer-facing chatbot is a typical signal for a transparency review. It is not automatic. It requires review.
Does Article 50 apply to every SME that uses ChatGPT?
No. Using ChatGPT does not classify the company by itself. You have to look at the use, the data and the potential role. Not every company that uses ChatGPT is an AI provider.
Did the Digital Omnibus postpone Article 50?
No. The Omnibus postponed certain Annex III high-risk duties. Article 50 transparency duties still apply from August 2026.
Is 2 December 2026 a new date for everyone?
No. It is not a generic new prohibitions package for every company. It is a narrow change. The calendar detail is in the AI Act guide and the AI legal risk assessment article.
Where do I start?
With the five questions on the AI Legal Assessment page, or write to [email protected]. If needed, the Assessment. If there is a provider contract, a contract review.
AI Act for companies: 2026–2028 guide →
Sources
- AI Act text, including Article 4: Regulation (EU) 2024/1689 (EUR-Lex).
- Spanish Data Protection Agency (AEPD): innovation and technology and the guide Adecuación al RGPD de tratamientos que incorporan Inteligencia Artificial.
- Spanish Artificial Intelligence Supervisory Agency: AESIA.
This page links those sources. It does not reproduce the statutory text and it does not classify a specific case.