# EU AI Act Article 50: Chatbots and AI labels

Source: https://aiexponent.com/eu-ai-act/article-50 · Content verified 2026-10-26

If people talk to your AI or see what it makes, you usually have to tell them.

**Does this apply to me?**

- Yes: Chatbot or generator provider (Art. 50(1))
- Yes: Deepfake or emotion AI deployer (Art. 50(3))
- Depends: Law-enforcement exceptions (Art. 50(1))

**By when?**

- 2 Aug 2026: Applies (Art. 113, second paragraph)
- 2 Dec 2026: Older generators: marking (Art. 111(4), as amended)

**What must I produce?**

- AI notice (Art. 50(1))
- Machine marking (Art. 50(2))
- Deepfake notice (Art. 50(4))
- Emotion AI notice (Art. 50(3))

Status: In force. Maximum fine: €15M or 3% (Art. 99(4), point (g)). 9 min read.

Plain-words summary. The quoted text below is the law.

Reviewed by Ajay Pundhir, founder (https://aiexponent.com/about#ajay-pundhir)

Article 50 duties sit on top of every other duty. A visible label does not meet the machine-readable marking duty in Article 50(2), and emotion recognition in workplaces and education is prohibited under Article 5(1)(f) unless used for medical or safety reasons.

## How it works: Which notice do I need?

_Rows are what your system does; columns are your role._

| | Provider | Deployer |
| --- | --- | --- |
| Talks or chats with people | Design it so people are told they are interacting with AI, unless that is obvious (Art. 50(1)) | Paragraph 1 puts this duty on the provider (Art. 50(1)) |
| Generates audio, images, video or text | Mark the outputs in a machine-readable format, detectable as AI-generated (Art. 50(2)) | Paragraph 2 puts this duty on the provider (Art. 50(2)) |
| Makes a deepfake image, audio or video | The marking duty in paragraph 2 applies to the generator (Art. 50(2)) | Disclose that it is AI-generated or manipulated; for artistic or creative work, a limited disclosure (Art. 50(4), first subparagraph) |
| Writes text published to inform the public | The marking duty in paragraph 2 applies to the generator (Art. 50(2)) | Disclose it, unless it had human review or editorial control and someone holds editorial responsibility (Art. 50(4), second subparagraph) |
| Recognises emotions or sorts people by biometric data | Paragraph 3 puts this duty on the deployer (Art. 50(3)) | Inform the people exposed, and process their data under the EU data protection rules (Art. 50(3)) |

_Each paragraph has exceptions, for example where the use is authorised by law to detect or prosecute criminal offences. Read the quoted text._

### Where each notice goes

1. **First chat message** (visible): Tell people they are interacting with an AI system, at the latest at the first interaction. (Art. 50(1))
2. **Inside the generated file** (not visible): A machine-readable mark that shows the content is AI-generated. (Art. 50(2))
3. **On the deepfake video** (visible): A disclosure that the content is AI-generated or manipulated. (Art. 50(4), first subparagraph)

_A made-up product screen. The mark in spot 2 sits in the file for software to detect; it is not a visible label._

## Who does what

**Provider**

- Design systems that interact with people so they are told it is AI, unless obvious. (Art. 50(1))
- Mark generated audio, image, video and text in a machine-readable format. (Art. 50(2))

**Deployer**

- Inform people exposed to emotion recognition or biometric categorisation. (Art. 50(3))
- Disclose deepfakes. (Art. 50(4), first subparagraph)
- Disclose AI-generated text published to inform the public, unless it had editorial control. (Art. 50(4), second subparagraph)

## What to do

- [ ] List your systems by what they do and your role for each, and check Chapter III too: Article 50 adds to those duties. (Art. 50(6))
- [ ] Tell people they are interacting with AI at the first interaction, unless that is obvious. (Art. 50(1))
  Worked example: A customer-service chatbot.
- [ ] Choose a machine-readable marking technique that is effective, interoperable, robust and reliable as far as technically feasible. (Art. 50(2))
- [ ] Disclose deepfakes; for evidently artistic, creative, satirical or fictional work, disclose in a way that does not hamper the work. (Art. 50(4), first subparagraph)
  Worked example: AI images in a marketing campaign.
- [ ] For AI-generated text published to inform the public: disclose it, unless it had human review or editorial control and a person holds editorial responsibility. (Art. 50(4), second subparagraph)
- [ ] Before using emotion recognition or biometric categorisation, check Article 5(1)(f) and Annex III; then inform the people exposed. (Art. 50(3))
- [ ] Give the information clearly and distinguishably, at the latest at the first interaction or exposure, and meet the accessibility requirements. (Art. 50(5))

Printable checklist: print https://aiexponent.com/eu-ai-act/article-50. Content verified 26 Oct 2026. Not legal advice. A planning aid; the quoted text of the Act is the law.

## Worked examples

**Example: A customer-service chatbot.** A software company builds a chatbot that answers customer questions and sells it to online shops.

- Outcome: As the provider, the company designs the chatbot so shoppers are told they are talking to an AI system, unless that is obvious, at the latest at the first interaction.
- Why: Paragraph 1 puts the design duty on providers; paragraph 5 sets when and how the information is given. (Art. 50(1), Art. 50(5))

**Example: AI images in a marketing campaign.** A marketing team uses an image generator to create a realistic photo of a real city street for an advert.

- Outcome: The image can count as a deep fake, so the team, as deployer, discloses that it is AI-generated. The generator’s provider must mark its outputs in a machine-readable format.
- Why: Article 3(60) covers realistic content that resembles existing places; paragraph 4 puts disclosure on deployers and paragraph 2 puts marking on providers. (Art. 3(60), Art. 50(4), first subparagraph, Art. 50(2))

## Common mistakes

- Common mistake: Reading the later marking date as the date for every generator.
  What the text says: “that have been placed on the market before 2 August 2026 shall take the necessary steps in order to comply with Article 50(2) by 2 December 2026” (Art. 111(4))
- Common mistake: Treating a visible AI label as the marking duty.
  What the text says: “marked in a machine-readable format and detectable as artificially generated or manipulated” (Art. 50(2))
- Common mistake: Calling every chatbot obviously AI to skip the notice.
  What the text says: “unless this is obvious from the point of view of a natural person who is reasonably well-informed, observant and circumspect” (Art. 50(1))
- Common mistake: Assuming a notice is enough for emotion recognition at work.
  What the text says: “to infer emotions of a natural person in the areas of workplace and education institutions” (Art. 5(1)(f))
- Common mistake: Treating Article 50 as the only duty of a limited-risk system.
  What the text says: “Paragraphs 1 to 4 shall not affect the requirements and obligations set out in Chapter III” (Art. 50(6))
- Common mistake: Quoting the top fine band, the one for prohibited practices.
  What the text says: “transparency obligations for providers and deployers pursuant to Article 50” (Art. 99(4)(g))

## What changed

Regulation (EU) 2026/1744, in force since 27 Jul 2026. Removed and context lines quote Regulation (EU) 2024/1689 as adopted; added lines quote the amending Regulation.

### Article 50(7)

Codes of practice on marking and labelling: assessed by the Commission instead of approved by implementing act

```diff
- 7. The AI Office shall encourage and facilitate the drawing up of codes of practice at Union level to facilitate the effective implementation of the obligations regarding the detection and labelling of artificially generated or manipulated content. The Commission may adopt implementing acts to approve those codes of practice in accordance with the procedure laid down in Article 56 (6). If it deems the code is not adequate, the Commission may adopt an implementing act specifying common rules for the implementation of those obligations in accordance with the examination procedure laid down in Article 98(2).
+ 7. The Commission shall encourage and facilitate the drawing up of codes of practice at Union level to facilitate the effective implementation of the obligations regarding the detection, marking and labelling of artificially generated or manipulated content. The Commission, taking utmost account of the opinion of the Board, shall assess whether adherence to those codes of practice is adequate to ensure compliance with the obligations laid down in paragraphs 2 and 4 of this Article, in accordance with the procedure laid down in Article 56(6). If it deems the code of practice to be inadequate, the Commission may adopt an implementing act specifying common rules for the implementation of those obligations in accordance with the examination procedure laid down in Article 98(2).
```

Source: Regulation (EU) 2026/1744, Article 1, point (20), https://eur-lex.europa.eu/eli/reg/2026/1744/oj/eng (verified 26 Oct 2026).

### Article 111(4)

Generative systems already on the market: Article 50(2) by 2 Dec 2026

```diff
+ 4. Providers of AI systems, including general-purpose AI systems, generating synthetic audio, image, video or text content, that have been placed on the market before 2 August 2026 shall take the necessary steps in order to comply with Article 50(2) by 2 December 2026.
```

Source: Regulation (EU) 2026/1744, Article 1, point (39), https://eur-lex.europa.eu/eli/reg/2026/1744/oj/eng (verified 4 Oct 2026).

## What Article 50 says

> **50(1)** 1. Providers shall ensure that AI systems intended to interact directly with natural persons are designed and developed in such a way that the natural persons concerned are informed that they are interacting with an AI system, unless this is obvious from the point of view of a natural person who is reasonably well-informed, observant and circumspect, taking into account the circumstances and the context of use. This obligation shall not apply to AI systems authorised by law to detect, prevent, investigate or prosecute criminal offences, subject to appropriate safeguards for the rights and freedoms of third parties, unless those systems are available for the public to report a criminal offence.

> **50(2)** 2. Providers of AI systems, including general-purpose AI systems, generating synthetic audio, image, video or text content, shall ensure that the outputs of the AI system are marked in a machine-readable format and detectable as artificially generated or manipulated. Providers shall ensure their technical solutions are effective, interoperable, robust and reliable as far as this is technically feasible, taking into account the specificities and limitations of various types of content, the costs of implementation and the generally acknowledged state of the art, as may be reflected in relevant technical standards. This obligation shall not apply to the extent the AI systems perform an assistive function for standard editing or do not substantially alter the input data provided by the deployer or the semantics thereof, or where authorised by law to detect, prevent, investigate or prosecute criminal offences.

> **50(3)** 3. Deployers of an emotion recognition system or a biometric categorisation system shall inform the natural persons exposed thereto of the operation of the system, and shall process the personal data in accordance with Regulations (EU) 2016/679 and (EU) 2018/1725 and Directive (EU) 2016/680, as applicable. This obligation shall not apply to AI systems used for biometric categorisation and emotion recognition, which are permitted by law to detect, prevent or investigate criminal offences, subject to appropriate safeguards for the rights and freedoms of third parties, and in accordance with Union law.

> **50(4), first subparagraph** 4. Deployers of an AI system that generates or manipulates image, audio or video content constituting a deep fake, shall disclose that the content has been artificially generated or manipulated. This obligation shall not apply where the use is authorised by law to detect, prevent, investigate or prosecute criminal offence. Where the content forms part of an evidently artistic, creative, satirical, fictional or analogous work or programme, the transparency obligations set out in this paragraph are limited to disclosure of the existence of such generated or manipulated content in an appropriate manner that does not hamper the display or enjoyment of the work.

> **50(4), second subparagraph** Deployers of an AI system that generates or manipulates text which is published with the purpose of informing the public on matters of public interest shall disclose that the text has been artificially generated or manipulated. This obligation shall not apply where the use is authorised by law to detect, prevent, investigate or prosecute criminal offences or where the AI-generated content has undergone a process of human review or editorial control and where a natural or legal person holds editorial responsibility for the publication of the content.

> **50(5)** 5. The information referred to in paragraphs 1 to 4 shall be provided to the natural persons concerned in a clear and distinguishable manner at the latest at the time of the first interaction or exposure. The information shall conform to the applicable accessibility requirements.

> **50(6)** 6. Paragraphs 1 to 4 shall not affect the requirements and obligations set out in Chapter III, and shall be without prejudice to other transparency obligations laid down in Union or national law for deployers of AI systems.

> **50(7), as amended by Regulation (EU) 2026/1744** 7. The Commission shall encourage and facilitate the drawing up of codes of practice at Union level to facilitate the effective implementation of the obligations regarding the detection, marking and labelling of artificially generated or manipulated content. The Commission, taking utmost account of the opinion of the Board, shall assess whether adherence to those codes of practice is adequate to ensure compliance with the obligations laid down in paragraphs 2 and 4 of this Article, in accordance with the procedure laid down in Article 56(6). If it deems the code of practice to be inadequate, the Commission may adopt an implementing act specifying common rules for the implementation of those obligations in accordance with the examination procedure laid down in Article 98(2).

> **111(4), as amended by Regulation (EU) 2026/1744** 4. Providers of AI systems, including general-purpose AI systems, generating synthetic audio, image, video or text content, that have been placed on the market before 2 August 2026 shall take the necessary steps in order to comply with Article 50(2) by 2 December 2026.

> **3(39)** (39) ‘emotion recognition system’ means an AI system for the purpose of identifying or inferring emotions or intentions of natural persons on the basis of their biometric data;

> **3(40)** (40) ‘biometric categorisation system’ means an AI system for the purpose of assigning natural persons to specific categories on the basis of their biometric data, unless it is ancillary to another commercial service and strictly necessary for objective technical reasons;

> **3(60)** (60) ‘deep fake’ means 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;

> **5(1)(f)** (f) the placing on the market, the putting into service for this specific purpose, or the use of AI systems to infer emotions of a natural person in the areas of workplace and education institutions, except where the use of the AI system is intended to be put in place or into the market for medical or safety reasons;

> **99(4)(g)** (g) transparency obligations for providers and deployers pursuant to Article 50.

Selected paragraphs, quoted exactly from Regulation (EU) 2024/1689: https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng (checked 26 Oct 2026).

## Questions about Article 50

### When does Article 50 apply?

It has applied since 2 Aug 2026 (Art. 113, second paragraph). Generative systems placed on the market before 2 Aug 2026 meet Art. 50(2) by 2 Dec 2026 (Art. 111(4), as amended).

### What is the maximum fine for breaching Article 50?

Up to €15 million or 3% of total worldwide annual turnover for the preceding financial year, whichever is higher (Art. 99(4), point (g)). For SMEs, including start-ups, the fine is capped at whichever of the two is lower (Art. 99(6)). Since 27 Jul 2026, the same lower cap applies to small mid-cap enterprises (Art. 99(6a)).

### Did the Digital Omnibus change Article 50?

Yes. Regulation (EU) 2026/1744, in force since 27 Jul 2026, makes these changes. Codes of practice on marking and labelling: assessed by the Commission instead of approved by implementing act. Generative systems already on the market: Article 50(2) by 2 Dec 2026. The section "What changed" quotes the old and new text.

### Is there an AiExponent tool for Article 50?

No. AiExponent has no released tool for Article 50.

### Must a chatbot tell people it is AI?

Yes, as a rule. Article 50(1) requires providers to design AI systems that interact directly with people so those people are informed they are interacting with an AI system, unless that is obvious to a reasonably well-informed, observant and circumspect person. Article 50(5) says the information comes at the latest at the first interaction.

### Do AI-generated images need a label?

Two duties apply. Providers of systems that generate images, audio, video or text must mark the outputs in a machine-readable format so they are detectable as AI-generated (Article 50(2)). Deployers must disclose that an image, audio or video deepfake is AI-generated or manipulated (Article 50(4)). A visible label alone does not meet the marking duty.

### What is a deepfake under the AI Act?

Article 3(60) defines a deep fake as 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. Text is not in the definition; AI-generated text published to inform the public has its own rule in Article 50(4).

## Further reading

- [Ajay Pundhir on the AI rule deadlines coming up, including Article 50(2) marking (AskAjay)](https://askajay.ai/thinking/regulation-note-2026-10?utm_source=aiexponent&utm_medium=explainer&utm_campaign=further_reading)
- [Ajay Pundhir on what deployers must do after the Omnibus (AskAjay)](https://askajay.ai/thinking/eu-ai-act-strategic-guide-business-leaders?utm_source=aiexponent&utm_medium=explainer&utm_campaign=further_reading)

---

Not legal advice. Not a notified body. The tools produce evidence, not conformity assessment.
All docs as Markdown: https://aiexponent.com/llms.txt · Guide for coding agents: https://aiexponent.com/agents.md