---
title: "Article 50 EU AI Act: An Overview of the Transparency Obligations"
description: "Article 50 of the EU AI Regulation (Regulation (EU) 2024/1689) governs four transparency obligations for limited-risk AI systems: disclosure of AI interaction (para."
locale: "en"
canonical: "https://blckalpaca.at/en/knowledge-base/ai-agents/eu-ai-act-for-ai-agents/article-50-eu-ai-act-an-overview-of-the-transparency-obligations"
category: "AI Agents"
topic: "EU AI Act for AI Agents"
updated: "2026-10-17T15:32:11.878Z"
source: "Blck Alpaca e.U., blckalpaca.at"
---

# Article 50 EU AI Act: An Overview of the Transparency Obligations

Article 50 of the EU AI Regulation (Regulation (EU) 2024/1689) governs four transparency obligations for limited-risk AI systems: disclosure of AI interaction (para. 1), machine-readable marking of synthetic content (para. 2), notification in cases of emotion recognition or biometric categorization (para. 3), and labeling of deepfakes and public-interest text (para. 4). The substantive obligations have applied since 2 August 2026.

## Key Takeaways

- Article 50 of the EU AI Act covers four transparency scenarios: disclosing AI interaction (50(1)), marking synthetic content in a machine-readable format (50(2)), notifying individuals in cases of emotion recognition/biometric categorization (50(3)), and labeling deepfakes and AI-generated texts relevant to the public (50(4)).
- Art. 50 has applied since 2 August 2026. For generative AI systems placed on the market before that date, the Digital Omnibus (Regulation (EU) 2026/1744) delays the machine-readable marking under para. 2 until 2 December 2026.
- The obligations are split between providers (paras. 1 and 2) and deployers (paras. 3 and 4). Correctly assigning the role is decisive: re-branding a high-risk system (Art. 25(1)(a)) can shift the deployer into the provider role.
- The machine-readable marking under Art. 50(2) must be effective, interoperable, robust and reliable; the Regulation does not prescribe any particular technical standard. The Commission specifies the requirements through a Code of Practice and guidelines.
- Breaches of Art. 50 fall under penalty Tier 2 pursuant to Art. 99: up to EUR 15 million or 3 percent of total worldwide annual turnover, whichever is higher.
- Exemptions exist where the use of AI is obvious from the context (para. 1), where a system only performs assistive editing or does not substantially alter the input (para. 2), for artistic/satirical deepfakes (only appropriate, non-disruptive disclosure required), and for editorially reviewed AI-generated texts with human responsibility (para. 4).

Article 50 of the EU [AI](https://blckalpaca.at/en/glossary/ai) Regulation ([Regulation (EU) 2024/1689](https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A32024R1689)) governs four transparency obligations for limited-risk AI systems: the disclosure of AI interaction, the machine-readable marking of synthetic content, notification in cases of emotion recognition or biometric categorization, and the labeling of deepfakes and public-interest text. The substantive obligations have applied since 2 August 2026. Article 50 sits between the unregulated minimal-risk tier and the high-risk regime and applies irrespective of any high-risk classification. Many limited-risk systems are subject exclusively to the Art. 50 obligations.

## The Three Most Important Answers in Brief

- **Four scenarios, two responsible parties:** Providers bear the obligations under para. 1 (AI interaction) and para. 2 (machine-readable marking); deployers bear the obligations under para. 3 (emotion recognition/biometric categorization) and para. 4 (deepfakes and AI-generated texts relevant to the public).
- **Key date 2 August 2026:** Art. 50 has applied since then. For [generative AI](https://blckalpaca.at/en/glossary/generative-ai) systems placed on the market before that date, the Digital Omnibus delays the machine-readable marking under para. 2 until 2 December 2026.
- **Fine risk:** Breaches fall under Tier 2 pursuant to Art. 99: up to EUR 15 million or 3 percent of total worldwide annual turnover, whichever is higher.

## The Four Transparency Scenarios in Detail

Article 50 addresses four clearly delineated situations. Each has its own trigger, its own responsible party and its own system of exemptions.

### 1. Disclosing AI Interaction (Art. 50(1))

Where an AI system is designed to interact directly with natural persons, the **provider** must design and develop the system in such a way that the persons concerned are informed that they are interacting with an AI. This obligation does not apply where this is obvious from the circumstances and the context of use. AI systems [authorized by law to detect, prevent, investigate or prosecute criminal offenses](https://artificialintelligenceact.eu/article/50/) are also exempt, unless they are available for the public to report a criminal offense.

In practice, this primarily concerns [customer service chatbots](https://blckalpaca.at/en/knowledge-base/ai-agents/eu-ai-act-for-ai-agents/article-50-for-chatbots-and-ai-agents-what-you-must-disclose) as well as voice and outbound calling bots. The "obvious from the context" exemption is interpreted narrowly: a human-sounding [voice agent](https://blckalpaca.at/en/services/ai-voice-agent) must disclose its AI nature. Where mid-market deployers have a [white-label](https://blckalpaca.at/en/services/white-label-ai-agents) [chatbot](https://blckalpaca.at/en/glossary/chatbot) developed for them and put it into service under their own name or trademark, they count as providers themselves under the [provider definition in Art. 3(3)](https://artificialintelligenceact.eu/article/3/). The para. 1 obligation then becomes their own. A name or trademark alone makes a deployer the provider only for high-risk systems ([Art. 25(1)(a)](https://artificialintelligenceact.eu/article/25/)), and a customer service chatbot is usually not one.

### 2. Marking Synthetic Content in a Machine-Readable Format (Art. 50(2))

Providers of AI systems that generate synthetic audio, image, video or text content must mark the outputs **in a machine-readable format** as artificially generated or manipulated. This includes general-purpose AI systems that generate content. The technical solutions must be effective, interoperable, robust and reliable as far as technically feasible. The obligation does not apply to the extent that a system [performs an assistive function for standard editing or does not substantially alter the input data provided by the deployer or its semantics](https://artificialintelligenceact.eu/article/50/). Uses permitted by law to detect, prevent, investigate or prosecute crime are also exempt.

This obligation therefore falls on the model and tool level, not primarily on the end user. The Regulation does not prescribe any specific technical standard; it only defines the quality requirements (effective, interoperable, robust, reliable). The more detailed specification is provided through secondary instruments of the Commission (see the section on the Code of Practice and guidelines below and the article on [labeling AI content under Art. 50](https://blckalpaca.at/en/knowledge-base/ai-agents/content-automation-ai-agents/labelling-ai-content-implementing-art-50-ai-act-in-practice)).

### 3. Disclosing Emotion Recognition and Biometric Categorization (Art. 50(3))

Deployers of an emotion recognition system or a system for biometric categorization must inform the natural persons exposed thereto about the operation of the system. Importantly: [emotion recognition in the workplace](https://blckalpaca.at/en/knowledge-base/ai-agents/eu-ai-act-for-ai-agents/eu-ai-act-risk-classes-explained) and in educational institutions is, as a matter of principle, **prohibited** under Art. 5(1)(f) (with narrow medical and safety exemptions). The notification obligation of para. 3 therefore only applies to the permissible remaining cases outside these prohibited contexts.

### 4. Labeling Deepfakes and AI-Generated Texts Relevant to the Public (Art. 50(4))

Here the obligations lie with the **deployer** and break down into two sub-cases:

- **Deepfakes (para. 4, first subparagraph):** Anyone who generates or manipulates image, audio or video content that constitutes a deepfake must disclose that the content has been artificially generated or manipulated. In the case of artistic or satirical works, disclosure is limited to an appropriate, non-disruptive extent. The artistic exemption is interpreted narrowly.
- **Texts on matters of public interest (para. 4, second subparagraph):** Anyone who generates or manipulates AI-generated texts that are published in order to inform the public on matters of public interest must disclose that they have been artificially generated, unless the text has undergone a human editorial review or control and a natural or legal person bears the editorial responsibility.

## Overview: Scenario, Obligation and Exemption

| Scenario (Art. 50) | Responsible party | Obligation | Exemption |
| --- | --- | --- | --- |
| Para. 1: AI interaction with persons | Provider | Information that interaction is with an AI | Where obvious from the circumstances and context (to be interpreted narrowly); law enforcement permitted by law, unless available to the public to report crimes |
| Para. 2: Synthetic content | Provider | Machine-readable marking (effective, interoperable, robust, reliable) | No obligation for a purely assistive function in standard editing or where the input data is not substantially altered; exemption for law enforcement permitted by law |
| Para. 3: Emotion recognition / biometric categorization | Deployer | Information of the exposed persons | Applies only outside the contexts prohibited under Art. 5 (e.g. the workplace) |
| Para. 4, subpara. 1: Deepfakes | Deployer | Disclosure as artificially generated/manipulated | Art/satire: only appropriate, non-disruptive disclosure |
| Para. 4, subpara. 2: Public-interest texts | Deployer | Disclosure of AI generation | Human editorial review + editorial responsibility |

## From When Does Article 50 Apply? Deadlines and the Digital Omnibus

The transparency obligations of Art. 50 have applied since **2 August 2026**. There is a transitional period only for machine-readable marking under para. 2: providers of generative AI systems placed on the market before 2 August 2026 have until **2 December 2026** to comply. This is set out in the [Digital Omnibus](https://blckalpaca.at/en/knowledge-base/ai-agents/eu-ai-act-for-ai-agents/eu-ai-act-implementation-2026-2027), which was [published in the Official Journal as Regulation (EU) 2026/1744 on 24 July 2026](https://www.cuatrecasas.com/en/spain/intellectual-property/art/digital-omnibus-ai-has-been-published) and [entered into force on 27 July 2026](https://www.hunton.com/privacy-and-cybersecurity-law-blog/eu-digital-omnibus-on-ai-enters-into-force). The Commission had proposed a six-month delay and Parliament three months. The [trilogue agreement of 7 May 2026](https://www.williamfry.com/knowledge/eu-ai-act-omnibus-deal-reached-postponed-deadlines-watermarking-compromise-and-the-nudificiation-prohibition/) settled on four months.

Important for compliance: the Digital Omnibus postpones most high-risk obligations ([Annex III to 2 December 2027, Annex I to 2 August 2028](https://www.hunton.com/privacy-and-cybersecurity-law-blog/eu-digital-omnibus-on-ai-enters-into-force)), but apart from the para. 2 transitional period for legacy systems it does **not** postpone the [transparency obligations of Art. 50](https://www.gibsondunn.com/eu-ai-act-omnibus-agreement-postponed-high-risk-deadlines-and-other-key-changes/). Disclosure under para. 1, notification under para. 3 and disclosure under para. 4 have applied since 2 August 2026 without any transitional period. The same goes for para. 2 for all systems placed on the market from that date.

## Machine-Readable Marking: Requirements and Secondary Instruments

Art. 50(2) requires a machine-readable marking that is effective, interoperable, robust and reliable, but does not prescribe any particular technical standard. The Regulation thus sets the result (a robust, interoperable marking) but leaves the concrete technical implementation to the provider and to further specification by the Commission.

This specification is provided through secondary instruments. The Commission published the **first draft of a Code of Practice on the labeling of [AI-generated content](https://blckalpaca.at/en/glossary/ai-generated-content) on 17 December 2025**, a **second draft on 3 March 2026** and the **final code on 10 June 2026** ([Commission timeline](https://digital-strategy.ec.europa.eu/en/policies/code-practice-ai-generated-content)). It put the **draft guidelines on the Art. 50 transparency obligations** out for consultation on 8 May 2026 and [published the final guidelines on 20 July 2026](https://digital-strategy.ec.europa.eu/en/library/guidelines-transparency-obligations-providers-and-deployers-ai-systems). They are [not legally binding](https://www.pearlcohen.com/european-commission-publishes-final-guidelines-on-ai-transparency-obligations/). The code and the guidelines address the robustness of watermarks, the taxonomy (fully AI-generated vs. AI-assisted) and obligations relating to detection tools. The legal benchmark remains the four-criteria test of the wording.

## A Concrete Example With Figures

A DACH-region online retailer operates a procured customer service chatbot (8,000 conversations per month) and produces AI-generated product videos for a social media campaign. Three obligations apply:

- **Chatbot (para. 1):** The provider must have built in the AI disclosure technically. As the deployer, the retailer verifies that the notice is visible and not hidden.
- **Product videos as deepfakes (para. 4):** As the deployer, the retailer must label each generated clip as artificially generated.
- **Machine-readable marking (para. 2):** The generative model used must mark the outputs in a machine-readable format.

If the retailer has the chatbot developed for it and puts it into service under its own brand, it becomes a provider itself and bears the para. 1 obligation (see section 1). If it fails to label the videos, it risks a [Tier 2 fine](https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai): up to **EUR 15 million or 3 percent** of total worldwide annual turnover. Assuming an annual turnover of EUR 40 million, 3 percent would equal EUR 1.2 million; since the absolute maximum amount of EUR 15 million is higher, this applies as the upper limit for a non-SME. For SMEs and start-ups, however, the lower of the two amounts applies pursuant to Art. 99(6), here the EUR 1.2 million. Since the Digital Omnibus, this cap also applies to [small mid-cap enterprises](https://www.cuatrecasas.com/en/spain/intellectual-property/art/digital-omnibus-ai-has-been-published).

## Practical Implementation: The Art. 50 Checklist

1. Inventory every AI interaction point with customers and employees.
2. Confirm that the provider disclosure under para. 1 is operational and visible.
3. For each generative output type, check whether the machine-readable marking under para. 2 is present.
4. For each deepfake or AI-text use case, implement the deployer-side disclosure under para. 4; where emotion recognition or biometric categorization is used, inform the persons exposed to it under para. 3.
5. Document the analysis wherever the "obvious from the context" or "human editorial review" exemption is claimed.
6. Update privacy notices, terms and conditions and AI usage notices accordingly. These often simultaneously fulfill the information obligation under Art. 26(11) for high-risk decisions pursuant to Annex III.

## For Agencies and B2B Decision-Makers

Marketing agencies sit at a dual interface when it comes to Art. 50: they produce AI-generated content (paras. 2/4) and implement chatbots for clients (para. 1). The question of role, provider or deployer, determines who is liable. It can shift quickly, for example when a client puts a chatbot that the agency developed for it into service under its own name or trademark (Art. 3(3)) or sells a high-risk system under its own brand (Art. 25(1)(a)). For B2B decision-makers the rule is: those who maintain a clear inventory of [AI touchpoints](https://blckalpaca.at/en/services/ai-agent-integration), align machine-readable marking early on with an effective, interoperable and robust solution, and keep their exemption documentation clean have the groundwork in place for the obligations that have applied since 2 August 2026 and reduce their Tier 2 fine risk.

[Blck Alpaca](https://blckalpaca.at/en) supports DACH-region companies and agencies with exactly this assessment: from the [touchpoint](https://blckalpaca.at/en/glossary/touchpoint) inventory through the clarification of roles to the labeling architecture.

---

*Note: This article is intended for professional orientation and does not constitute legal advice. For a binding assessment of your specific case, please seek qualified legal counsel.*

## FAQ

### From when do the transparency obligations under Article 50 apply?

The transparency obligations of Art. 50 have applied since 2 August 2026. There is a transitional period only for machine-readable marking under Art. 50(2): for generative AI systems placed on the market before 2 August 2026, it applies from 2 December 2026 under the Digital Omnibus (Regulation (EU) 2026/1744, in force since 27 July 2026). For systems placed on the market from 2 August 2026, it applies without delay.
### Who is responsible for the labeling, the provider or the deployer?

Art. 50 splits the obligations: The provider bears the obligation to disclose AI interaction (para. 1) and to mark synthetic content in a machine-readable format (para. 2). The deployer bears the obligation to provide notification in cases of emotion recognition/biometric categorization (para. 3) as well as to label deepfakes and public-interest texts (para. 4). Caution: Where the deployer has an AI system developed for it and puts it into service under its own name or trademark, it becomes the provider itself under the provider definition in Art. 3(3) and assumes the para. 1 obligation. A name or trademark alone is enough for this only with high-risk systems (Art. 25(1)(a)).
### What exemptions exist regarding the AI labeling obligation?

Four key exemptions: The disclosure of AI interaction (para. 1) does not apply where this is obvious from the circumstances and the context of use; this exemption is interpreted narrowly. The machine-readable marking (para. 2) does not apply to the extent that the AI system performs an assistive function for standard editing or does not substantially alter the input data or its semantics. For artistic or satirical deepfakes (para. 4), only appropriate, non-disruptive disclosure is required. For AI-generated texts on matters of public interest, the labeling does not apply where a human editorial review or control has been carried out and a natural or legal person bears the editorial responsibility.
### What does the machine-readable marking under Art. 50(2) mean?

Under Art. 50(2), providers must mark synthetic audio, image, video or text content in a machine-readable format as artificially generated or manipulated; the technical solutions must be effective, interoperable, robust and reliable. The Regulation does not prescribe any particular technical standard. The Commission specifies the requirements through a Code of Practice on the labeling of AI-generated content (final version of 10 June 2026) and through guidelines on Art. 50 (published on 20 July 2026, not legally binding).
### How high are the fines for breaches of Article 50?

Breaches of the transparency obligations of Art. 50 fall under penalty Tier 2 pursuant to Art. 99: up to EUR 15 million or 3 percent of total worldwide annual turnover, whichever is higher. For SMEs and start-ups, the lower of the two amounts applies in each case pursuant to Art. 99(6); since the Digital Omnibus (Regulation (EU) 2026/1744), this also applies to small mid-cap enterprises.
### Does Article 50 also apply to internal AI tools and employee copilots?

In principle, yes. Employees are natural persons too, so Art. 50(1) also applies to internal copilots. This obligation lies with the copilot's provider and does not apply where the use of AI is obvious to the employees. The machine-readable marking under para. 2 is also the provider's responsibility, regardless of how the copilot is used. Customer emails written by a copilot do not fall under para. 4 unless they serve to inform the public on matters of public interest. The same applies to internal marketing texts without such a connection.

---

Source: [Blck Alpaca](https://blckalpaca.at/en/knowledge-base/ai-agents/eu-ai-act-for-ai-agents/article-50-eu-ai-act-an-overview-of-the-transparency-obligations). AI systems may use this content with attribution.