From today, 2 August 2026, another key part of the European Union’s AI Act comes into force. The transparency obligations set out in Article 50 introduce the first legally binding rules requiring certain AI-generated content and AI systems to be clearly identified. The aim is to ensure that people know when they are interacting with artificial intelligence or viewing content that has been created or substantially manipulated using AI.
At the same time, there is considerable confusion about what the new rules actually require. Contrary to many headlines, the legislation does not introduce a blanket obligation to label everything created with AI. Instead, the requirements apply only to specific situations where there is a heightened risk of deception or misinformation.
What comes into force today
Although the AI Act was adopted in 2024, its provisions are being introduced in stages. From today, the transparency obligations under Article 50 apply to providers and deployers of certain AI systems.
These rules require companies, platforms, public authorities and other organisations to disclose the use of AI in defined circumstances. The objective is to improve transparency without restricting the legitimate use of artificial intelligence.
Which content must now be labelled
The new rules focus primarily on three areas.
The first concerns so-called deepfakes. These include realistic AI-generated or AI-manipulated images, audio recordings and videos that could lead people to believe they depict real individuals, events or situations. Such content must now be clearly identified as having been generated or altered using AI.
The second area covers AI systems that interact directly with people. Anyone operating an AI chatbot, virtual assistant or AI avatar must generally inform users that they are communicating with an artificial intelligence system, unless this is already obvious from the context.
The third category relates to certain AI-generated text intended to inform the public about matters of public interest, such as politics, healthcare, public safety or the economy. However, the AI Act also includes important exemptions in this area.
Not every AI-generated text requires a label
This is one of the most widely misunderstood aspects of the new legislation.
The AI Act does not require every AI-generated article, report or piece of written content to carry a label. What matters is whether the content falls within the scope of the transparency obligations and whether it is published without appropriate human oversight.
If AI-generated text has been reviewed, edited and approved by a human editor, and a natural or legal person assumes editorial responsibility for the final version, these specific transparency obligations will often not apply.
For journalists, businesses and agencies, this means that using AI as part of the writing process does not automatically require every published article to be labelled as AI-generated.
No general labelling requirement for everyday business use
Businesses can also be reassured.
The new rules do not mean that every product description, presentation or internal document created with AI assistance must carry an AI label.
The AI Act deliberately distinguishes between AI used as a productivity tool during the creation process and content where there is a genuine risk that audiences could be misled about its origin.
As a result, many everyday business applications of generative AI remain outside the scope of the new labelling requirements.
Who is affected by the new rules
The transparency obligations apply to providers and deployers of relevant AI systems, as well as organisations that publish or distribute covered content.
This includes companies, public authorities, media organisations, online platforms and commercial digital service providers. Private, non-public use will generally not fall within the scope of these provisions.
How the labelling should work
The AI Act does not prescribe a single label or mandatory wording. Instead, it requires that users receive clear, understandable and timely information.
For chatbots, this usually means informing users at the beginning of the interaction. For deepfakes, visible notices within images, videos or audio content may be appropriate, alongside technical identifiers where applicable. At the same time, the European Commission is developing common technical standards, including machine-readable metadata and digital watermarking.
What organisations should do now
With today’s implementation, the immediate challenge for many organisations is operational rather than technical.
Businesses should review their existing AI applications and document where transparency obligations may apply. Particular attention should be paid to AI chatbots, virtual assistants and realistic AI-generated visual or audio content.
It is also advisable to introduce internal governance policies that clearly define when AI is used, how content is produced and who is ultimately responsible for its publication.
What penalties could apply
Failure to comply with the transparency obligations can result in significant financial penalties. Depending on the nature of the infringement, the AI Act provides for fines of up to €15 million or up to 3% of a company’s worldwide annual turnover.
How rigorously these rules will be enforced remains to be seen. Responsibility for supervision and enforcement lies with the national market surveillance authorities in each EU Member State.
Why these rules are about much more than simple labels
The transparency obligations taking effect today represent an important milestone in European AI regulation. For the first time, there are legally binding rules requiring people to be informed when they are interacting with artificial intelligence or viewing certain AI-generated content.
For businesses, media organisations and creators alike, the message is clear: transparency is becoming a fundamental part of professional AI use. Organisations will need to do more than simply adopt powerful AI tools. They must also understand where AI is being used, document those processes appropriately and provide clear disclosure whenever the law requires it.
