EU AI Act Article 50: Transparency Requirements Explained

On 2 August 2026, another important phase of the EU AI Act comes into effect. From this date, the transparency obligations set out in Article 50 become applicable.

For organizations, this means taking a closer look at where and how AI systems are used and determining whether users, customers, or the public must be informed about that use.

This does not mean that every use of AI must be labeled. Whether transparency obligations apply depends on the specific use case, the organization's role, and the type of content being generated or processed.

Key Takeaways

Transparency Requirements for Certain AI Systems

From August 2026, the transparency obligations under Article 50 apply to a range of AI systems, including those that interact directly with people, generative AI systems, and certain AI-generated or AI-manipulated content.

Labeling AI-Generated Content

Providers of certain AI systems that generate synthetic text, images, audio, or video must ensure that these outputs can be identified in a machine-readable format as AI-generated or AI-manipulated.

Disclosure Requirements for Published Content

For some types of content, machine-readable labeling alone is not sufficient.

Organizations that publish deepfakes or certain AI-generated or AI-manipulated content relating to matters of public interest may also be required to disclose to users that the content was created or altered using AI.

Transparency When People Interact with AI

Organizations must generally inform users when they are interacting directly with an AI system, unless this is already obvious from the context.

Where Should Organizations Start?

As part of preparing for compliance with the EU AI Act, organizations should first identify where AI technologies are being used across their business processes and assess whether any transparency obligations apply.

Key questions include:

• Where do we use AI systems that interact directly with customers, employees, or other individuals?

• Do our AI systems generate or modify text, images, audio, or video?

• Are we the provider of the AI system, or are we using a third-party solution?

• Do our AI-generated outputs require machine-readable labeling?

• Are we required to disclose to users or the public that content has been generated or manipulated using AI?

• What should these labels or disclosures look like in practice?

For many organizations, the last question is especially important. Once it has been established that a particular AI use case falls within the scope of Article 50, the next step is determining how the transparency requirement should be implemented.

Article 50: When Do Transparency Requirements Apply?

Article 50 establishes transparency obligations for providers and deployers of certain AI systems. The most relevant scenarios include the following.

1. People Interact Directly with an AI System

Individuals must generally be informed when they are interacting with an AI system, unless this is already obvious.

Typical examples include AI-powered chatbots and virtual assistants.

2. AI Systems Generate or Manipulate Content

Providers of AI systems that generate synthetic text, images, audio, or video must ensure that this content is identifiable in a machine-readable format as AI-generated or AI-manipulated.

This requirement focuses on technical labeling rather than informing end users directly.

3. Deepfakes Are Published

Organizations that create or publish AI-generated or AI-manipulated deepfakes must generally disclose that the content has been artificially generated or manipulated.

Unlike machine-readable labeling, this obligation requires information to be provided directly to people.

4. AI-Generated Content on Matters of Public Interest

Disclosure obligations may also apply to AI-generated or AI-manipulated text published to inform the public about matters of public interest.

However, the AI Act provides an important exception where the content has undergone meaningful human review or editorial oversight and a natural or legal person assumes editorial responsibility.

5. Emotion Recognition and Biometric Categorization

Organizations using certain emotion recognition or biometric categorization systems must generally inform affected individuals that these systems are being used.

Labeling vs. Disclosure: What's the Difference?

One of the most important distinctions for organizations is the difference between machine-readable labeling and disclosure. Machine-readable labeling enables systems and software to recognize that content has been generated or manipulated by AI through technical metadata or similar mechanisms.

Disclosure, on the other hand, means informing people directly that content has been created or altered using AI.

While Article 50 specifies when these obligations apply, organizations must also determine how they can meet them in practice.

How Can Organizations Implement the Transparency Requirements?

To support implementation, the Code of Practice on Transparency of AI-Generated Content, developed with the support of the European Commission and the AI Office, provides practical guidance for providers and deployers of generative AI systems (Code of Practice on the Transparency of AI-Generated Content).

Participation in the Code is voluntary. Organizations remain free to demonstrate compliance through other appropriate methods.

The Code outlines practical approaches for implementing machine-readable labeling of AI-generated content as well as visible disclosures that inform users when content has been created or manipulated using AI. In July 2026, the European Commission concluded that the Code provides an appropriate framework for supporting compliance with the obligations set out in Article 50(2), (4), and (5).

In addition, the European Commission published Guidelines on the Transparency Obligations under Article 50 in July 2026. These Guidelines are intended to help organizations and public authorities understand the scope of the legal requirements, while the Code of Practice focuses on practical implementation of labeling and disclosure measures (Guidelines on the transparency obligations under Article 50).