EU AI Act: what Article 50 requires of chatbots and AI-generated content
Since August 2, 2026, Article 50 of Regulation (EU) 2024/1689, known as the AI Act, has applied in the European Union. It requires informing people when they are talking to an AI and identifying content generated or manipulated with AI. The rule also reaches companies outside the EU, such as Ecuadorian ones, when their system’s output is used in the Union (European Commission, 2026).
We have already covered the law’s overall timeline and the situation in Ecuador (Wasivia, 2026). Here we focus on the transparency rules, which are already enforceable.
What Article 50 requires
The law distinguishes the provider, which develops the system and offers it under its own name, from the deployer, which uses it in its professional activity. A company can be both (AI Act, Art. 50).
| Obligation | Who is responsible | What it involves |
|---|---|---|
| Disclose that people are interacting with an AI (Art. 50(1)) | Provider | Chatbots, agents and avatars must disclose it from the start of the first interaction, unless it is obvious. |
| Mark synthetic content (Art. 50(2)) | Provider | AI-generated audio, images, video or text must carry a machine-readable marking and be detectable. |
| Disclose emotion recognition or biometric categorization (Art. 50(3)) | Deployer | People exposed to the system must be informed that it is operating. |
| Disclose deepfakes (Art. 50(4)) | Deployer | A visible or audible notice, at the latest at the first exposure. |
| Identify texts on matters of public interest (Art. 50(4)) | Deployer | State that the text was generated by AI, unless it has undergone human review or editorial control. |
The information must be provided in a clear and distinguishable manner and meet accessibility requirements. The Commission says the “obvious” exception “should be interpreted in a restrictive manner” (European Commission, 2026).
There are important nuances:
- Marking doesn’t cover everything. Standard assisted editing, source code and outputs that only pass between machines are excluded.
- A technical marking is not enough for a deepfake. It needs a notice that anyone can perceive. For artistic or satirical works, a notice that doesn’t hamper the work is enough.
- Spell-checking is not editorial review. The exception requires a person to examine the substance and someone to take editorial responsibility.
Deadlines, tools and penalties
There is only one extra deadline: generative systems placed on the market before August 2, 2026, have until December 2, 2026, to comply with the marking requirement in Article 50(2). It was set by the new Article 111(4), added by the Digital Omnibus on AI (AI Act, Art. 111). Everything else has applied since August, and content created earlier does not need to be labeled retroactively.
The Commission has published guidelines on Article 50 (European Commission, 2026). There is also a code of practice on the transparency of AI-generated content, whose final version was published on June 10, 2026. It is voluntary, and the Commission and the European AI Board consider it adequate for demonstrating compliance. By the end of July, about 190 organizations had signed it. The EU also provides icons for labeling AI-generated content (European Commission, 2026).
Breaching Article 50 can cost up to EUR 15 million or 3% of total worldwide annual turnover, whichever is higher. For SMEs, whichever is lower applies (AI Act, Art. 99). Oversight falls mainly to national market surveillance authorities.
What it means for companies in Ecuador
The fact that Ecuador has no AI law does not exempt you from the European one. It reaches companies in third countries when their system’s output is used in the EU (AI Act, Art. 2).
Some typical cases:
- A travel agency whose virtual assistant handles bookings from Spain.
- A software company that sells a platform in Europe that generates images or text.
- An exporter running campaigns in the EU that use synthetic voices or faces.
If nothing your AI produces is used in the EU, Article 50 does not apply to you, but telling users they are dealing with an AI is still good practice.
What to do now
- Take inventory. Identify which AI systems serve people in the EU or produce content used there, and whether you are a provider, a deployer or both.
- Add the notice to your chatbot. Show it at the start, in plain language and an accessible format.
- Review marking with your vendors. Ask what marking they apply, whether they signed the code of practice and how the marking is preserved in your workflows.
- Label what people see. Define when to use the EU icons on deepfakes and texts on matters of public interest.
- Formalize editorial review. If you publish informational texts with AI, document who reviews the substance and who is accountable for the publication.
- Mind the December deadline. If you offer a generative system placed on the market before August, the marking deadline is December 2, 2026.
At Wasivia, we design generative AI and RAG solutions with transparency built in from the start. We also help you organize your governance through our AI consulting and strategy.
This article is for informational purposes only and is not legal advice. For your specific case, consult a specialized lawyer.
References
- European Commission — Transparency obligations under Article 50 of the AI Act: questions and answers (Jul 24, 2026)
- European Commission — Guidelines on the transparency obligations under Article 50
- European Commission — Code of practice on the transparency of AI-generated content
- AI Act Service Desk — AI Act, Article 50: transparency obligations
- AI Act Service Desk — AI Act, Article 2: scope
- AI Act Service Desk — AI Act, Article 99: penalties
- AI Act Service Desk — AI Act, Article 111: transitional provisions (version including the Digital Omnibus on AI)
- Wasivia — AI regulation in 2026: what changes for companies (EU and Ecuador)
