EU AI Act: Article 50 Transparency Rules Now Apply

On 2 August 2026, the next major set of obligations under the EU Artificial Intelligence Act became applicable. The most immediate change for businesses is Article 50, which introduces transparency requirements for certain AI systems.

These rules apply to specific uses of interactive AI, generative AI, emotion recognition, biometric categorisation, deepfakes, and AI-generated or manipulated text published on matters of public interest.

At the same time, the AI Omnibus has extended the application dates for certain high-risk AI requirements. That does not postpone Article 50. The transparency requirements apply now.

TL;DR

Article 50 transparency obligations apply from 2 August 2026.
AI systems that directly interact with people generally have to disclose that the user is interacting with AI, unless this is already obvious from the circumstances.
Providers of systems generating synthetic audio, images, video, or text must use machine-readable markings that allow the content to be detected as AI-generated or manipulated, subject to specific exceptions.
Deployers must inform people when they are exposed to emotion recognition or biometric categorisation systems.
Deepfakes must be disclosed, subject to specific exceptions.
AI-generated or manipulated text published to inform the public about matters of public interest must be disclosed unless it has undergone human review or editorial control and has appropriate editorial responsibility.
A limited transition until 2 December 2026 applies to the Article 50(2) marking requirement for AI systems placed on the market before 2 August 2026.
Article 50 violations can result in fines of up to €15 million or 3% of worldwide annual turnover, subject to the AI Act’s rules on proportionality and SMEs.
Enforcement is primarily carried out by national market-surveillance authorities. The AI Office has specific enforcement responsibilities for systems within its remit.

What Changed on 2 August 2026?

Article 50 Transparency Obligations

Article 50 establishes transparency requirements for certain AI systems regardless of whether the system is classified as high-risk.

The obligations cover four main situations:

AI systems that directly interact with people
AI systems generating synthetic audio, images, video, or text
Emotion recognition and biometric categorisation systems
Deepfakes and certain AI-generated or manipulated text published on matters of public interest

A business can fall under more than one of these requirements depending on how its AI system is used.

Interactive AI Systems

Providers of AI systems designed to interact directly with people must ensure that users are informed that they are interacting with an AI system.

There is an exception where this is obvious to a reasonably well-informed, observant, and circumspect person, taking the circumstances and context of use into account. The European Commission’s guidelines provide further detail on how this assessment should be made.

For a customer-facing chatbot, virtual assistant, or similar AI agent, the practical requirement is straightforward: users should not be left with the impression that they are communicating with a human when the system falls within Article 50(1).

Synthetic AI-Generated Content

Providers of AI systems that generate synthetic audio, images, video, or text must ensure that their outputs are marked in a machine-readable format and can be detected as artificially generated or manipulated.

The technical marking must be effective, interoperable, robust, and reliable to the extent technically feasible, taking into account the characteristics of the content, implementation costs, and the state of the art. The AI Act also contains exceptions for certain assistive editing functions and systems that do not substantially alter the input or its semantics.

This requirement is primarily a provider obligation. It is therefore particularly relevant to companies developing, supplying, or placing generative AI systems on the EU market.

Limited transition period

A limited transition period applies to AI systems placed on the market before 2 August 2026. For those systems, the Article 50(2) marking and detection obligation applies from 2 December 2026.

This is not a general four-month grace period for all Article 50 obligations. The transition applies specifically to the marking and detection requirement for qualifying existing systems.

Content generated before 2 August 2026 does not have to be labelled retroactively under Article 50, although the Commission encourages labelling where possible.

Emotion Recognition and Biometric Categorisation

Deployers of emotion recognition or biometric categorisation systems must inform the people exposed to those systems that they are being used.

The requirement applies to exposure in the relevant circumstances covered by Article 50(3). Other EU rules, including applicable data-protection legislation, continue to apply separately.

For businesses using these systems, compliance therefore involves more than deploying the technology itself. The people affected by the system must receive the required information.

Deepfakes and Public-Interest Text

Article 50 also covers certain AI-generated or manipulated content made available to the public.

Deepfakes

Deployers must disclose when image, audio, or video content constitutes a deepfake. The AI Act defines the relevant type of deepfake as AI-generated or manipulated content that resembles existing persons, objects, places, entities, or events and could falsely appear authentic or truthful.

There are specific exceptions. For example, where content forms part of an evidently artistic, creative, satirical, fictional, or similar work, the disclosure requirement is limited in how it must be presented.

AI-generated text on matters of public interest

Deployers must also disclose AI-generated or manipulated text published with the purpose of informing the public on matters of public interest. The obligation does not apply where the publication has undergone human review or editorial control and a person holds editorial responsibility for the publication. The human-review exception is more demanding than proofreading.

The European Commission states that human review involves deliberate examination of the substance by people with relevant knowledge and professional judgement. Editorial control involves a responsible editorial entity with authority to approve, modify, or reject the content on substantive grounds, including fact-checking and assessing source reliability. A superficial check, spell-check, or grammatical correction is not enough.

Enforcement and Penalties

Who Enforces Article 50?

Enforcement is primarily the responsibility of national competent market-surveillance authorities. The AI Office has a more limited role under Article 50. Its remit covers specific AI systems within its supervisory competence, including certain systems built on general-purpose AI models and certain systems integrated into very large online platforms or search engines.

The European Data Protection Supervisor is responsible for relevant AI systems used by EU institutions, bodies, and agencies. It is therefore inaccurate to describe the AI Office as the general enforcement authority for every business using AI in the EU.

Fines

For violations of transparency obligations under Article 50, administrative fines can reach: €15 million, or 3% of the undertaking’s total worldwide annual turnover for the preceding financial year, whichever is higher.

The AI Act requires relevant circumstances and proportionality to be considered when determining whether to impose a fine and its amount. Special limits also apply to SMEs and start-ups.

The higher €35 million or 7% ceiling applies to prohibited AI practices under Article 5. It is not the penalty level for ordinary Article 50 transparency violations.

Supplying incorrect, incomplete, or misleading information to notified bodies or national competent authorities can result in fines of up to €7.5 million or 1% of worldwide annual turnover, subject to the applicable rules.

For providers of general-purpose AI models, the Commission can impose fines of up to €15 million or 3% of worldwide annual turnover for relevant violations under Article 101.

The AI Omnibus: What Was Actually Delayed?

The Digital Omnibus on AI, Regulation (EU) 2026/1744, entered into force on 27 July 2026. It changed the application timetable for certain high-risk AI requirements. It did not remove or generally postpone Article 50 transparency obligations.

High-Risk AI Systems

The revised timetable is:

Category	Application date
Certain high-risk AI systems under Annex III	2 December 2027
High-risk AI systems embedded in regulated products under Annex I	2 August 2028
Article 50 transparency requirements	2 August 2026

The delay gives businesses additional time to prepare for the high-risk regime. It does not mean businesses can ignore the AI Act until 2027 or 2028. Other provisions are already applicable, including the prohibition rules and AI literacy requirements that started applying in February 2025, as well as the governance and GPAI obligations that became applicable in August 2025.

New Prohibited Practices

The AI Omnibus also introduced prohibitions covering AI systems used to generate non-consensual sexually explicit or intimate content and child sexual abuse material.

These additions form part of the revised AI Act framework and should not be confused with the transparency obligations that became applicable on 2 August 2026.

What This Means for Your Business

You use a chatbot or AI assistant

Check whether the system clearly informs users that they are interacting with AI. Do not rely on the assumption that users will automatically recognise the system as artificial. The Article 50 assessment depends on the system, its presentation, and the circumstances in which it is used.

You generate content with AI

Identify which systems you use and whether you are acting as a provider or deployer. For systems within Article 50(2), providers need a mechanism for machine-readable marking and detection of synthetic audio, image, video, and text content, subject to the relevant exceptions and transition rules. For published deepfakes, ensure the required disclosure is visible and appropriate to the content.

You publish AI-generated public-interest content

Determine whether the content is intended to inform the public on a matter of public interest. If it falls within Article 50(4), establish whether it has undergone genuine human review or editorial control. A basic proofreading process is not enough.

Businesses publishing this type of content should also keep evidence of their review and editorial processes so they can demonstrate how the exception was applied.

You use emotion recognition or biometric categorisation

Review where and how people are exposed to these systems. Where Article 50(3) applies, the people concerned must be informed. Data-protection requirements may also apply independently.

You develop or provide general-purpose AI

The AI Office has direct supervisory and enforcement responsibilities for providers of general-purpose AI models. The Commission can request documentation and information, conduct evaluations, require corrective measures, and impose fines within the limits set by the AI Act.

The Code of Practice on Transparency

The European Commission published the Code of Practice on Transparency of AI-generated Content in June 2026. The code is voluntary. It provides a practical way for providers and deployers to structure compliance with the marking and labelling obligations covered by Article 50. Signing the code does not replace the legal obligations.

Organisations that do not sign it remain responsible for demonstrating compliance through appropriate measures. On 31 July 2026, the Commission announced that more than 180 organisations had signed the code. The code should therefore be treated as a compliance tool, not as a substitute for the regulation itself.

Timeline

1 August 2024: EU AI Act enters into force
2 February 2025: Prohibitions and AI literacy requirements start applying
2 August 2025: Governance rules and GPAI obligations start applying
10 June 2026: Code of Practice on Transparency of AI-generated Content published
20 July 2026: Commission guidelines on Article 50 published
24 July 2026: Digital Omnibus on AI published in the Official Journal
27 July 2026: Digital Omnibus enters into force
2 August 2026: Article 50 transparency obligations apply
2 December 2026: Transition ends for Article 50(2) marking obligations for qualifying AI systems placed on the market before 2 August 2026
2 December 2027: Certain Annex III high-risk AI requirements apply
2 August 2028: Certain Annex I product-embedded high-risk AI requirements apply

Key Takeaways

Article 50 is already applicable. The date was 2 August 2026.
The rules cover specific forms of AI interaction, synthetic content, biometric and emotion-recognition systems, deepfakes, and certain public-interest text.
The limited December 2026 transition applies only to the Article 50(2) marking requirement for qualifying systems already placed on the market before 2 August 2026.
High-risk AI deadlines were extended, but Article 50 was not generally postponed.
Article 50 violations can result in fines of up to €15 million or 3% of worldwide annual turnover.
Enforcement is primarily handled by national market-surveillance authorities, with the AI Office responsible for systems within its specific remit.
Human review of AI-generated public-interest text must be substantive. Proofreading alone is not sufficient.
The Code of Practice is voluntary. The underlying Article 50 obligations are not.

Conclusion

The EU AI Act is not entering a single new compliance phase on one date. Different parts of the regulation have different application dates.

For businesses, the immediate issue is Article 50. Since 2 August 2026, certain AI systems must meet specific transparency requirements. The fact that some high-risk AI requirements have moved to 2027 and 2028 does not remove these obligations.

Businesses using customer-facing AI, generative AI, biometric categorisation, emotion recognition, or AI-generated public-interest content should identify which Article 50 obligations apply to their systems and document how those obligations are being met.

The European Commission has now published both the Article 50 guidelines and the Code of Practice, giving businesses concrete compliance material to work from.

Resources

Need help navigating the EU AI Act? Playful Sparkle has been engineering digital products since 2004, helping businesses assess AI implementations and meet technical transparency requirements. Our team can help you audit AI systems, identify applicable Article 50 obligations, implement transparency measures, and build compliant AI-powered digital products. Contact us to discuss your requirements.

EU AI Act: Article 50 Transparency Rules Now Apply

TL;DR