AI transparency: key business obligations under article 50 of the AI Act
The European Commission has finalized its consultation on guidelines concerning the transparency obligations set out in Article 50 of the AI Act. These provisions have applied since 2 August 2026 and cover requirements such as informing individuals when they interact with AI systems, labeling certain AI-generated or AI-manipulated content, and complying with obligations relating to deepfakes, emotion recognition systems, and biometric categorization systems.
Who is subject to the new obligations?
The guidelines apply to both:
- AI providers, and
- AI deployers.
The scope of obligations varies depending on the role of the organization and the type of AI system or content involved.
Key obligations under article 50 of the AI Act
Code of practice for AI content labeling
In June 2026, the European Commission published a Code of Practice on the labeling and disclosure of AI-generated content. The Code is intended to support organizations in meeting their obligations under Article 50 of the AI Act. Participation in the Code is voluntary. However, applying its recommended measures may help organizations demonstrate compliance with obligations relating to the labeling and disclosure of AI-generated or AI-modified content.
Why is this important for businesses?
According to the European Commission, the rapid development of generative AI makes it increasingly difficult to distinguish between human-created content and content generated by algorithms.
This increases the risk of:
- disinformation,
- manipulation,
- fraud, and
- impersonation.
The transparency requirements are intended to build trust and enable users to make informed judgments about the content they consume.
What should organizations do?
- assess whether they use AI systems covered by Article 50 of the AI Act,
- implement notices informing users about the use of AI,
- establish mechanisms for labeling AI-generated content,
- develop procedures for publishing deepfake content,
- review marketing, HR, customer service, and communication activities in light of the new transparency obligations.
Summary
Article 50 of the AI Act does not prohibit the use of artificial intelligence. Instead, it requires greater transparency. Organizations should ensure that users understand when they are interacting with AI and when they are consuming content generated or modified by artificial intelligence. For many businesses, this will require implementing new procedures, disclosures, and labeling mechanisms before any supervisory authority inspections take place.
Need support with AI Governance?
Find out how Crowe Poland’s experts can help your organization.
Frequently asked questions (FAQ)
Not necessarily. The obligation applies where users may not reasonably realize that they are communicating with an AI system. If the nature of the tool is obvious, additional notices may not be required.
Not every marketing asset created using AI must be visibly labeled as AI-generated content. Each case should be assessed individually by considering the type of content generated or manipulated, the extent of AI involvement, and whether the content falls into one of the categories covered by Article 50 of the AI Act, such as deepfakes.
No. Certain obligations also apply to organizations using AI in their operations, including marketing, HR, customer service, and corporate communications.
Typically, implementation will require cooperation between compliance, legal, IT, cybersecurity, marketing, HR, and customer service teams. AI transparency is becoming a cross-functional issue affecting the entire organization.
Not automatically. The applicability of Article 50 depends on the type of AI system and how it is used. Regardless of Article 50, other AI Act obligations may apply, particularly those relating to high-risk AI systems, as well as requirements under the GDPR.
Not always. Information should be provided in a clear, understandable, and context-appropriate manner. In many situations, notices will need to be displayed directly alongside the relevant service or content.
Yes. The AI Act provides for fines of up to EUR 15 million or, for businesses, up to 3% of total worldwide annual turnover from the previous financial year for certain breaches. The level of sanctions depends on the nature of the infringement and the specific circumstances of the case.
Yes. It is considered good practice to document internal procedures, publication rules, and measures adopted to fulfill transparency obligations. Such documentation may prove valuable during inspections or regulatory investigations.
The best starting point is an inventory of AI tools used across the organization. Only after identifying where and how AI is being used can a company effectively implement the required transparency measures, disclosures, and procedures.
See also:
