As of 2 August 2026, the new transparency rules of the AI Act will apply in the European Union. The goal is simple: citizens need to know when they are interacting with AI and when the content they are watching, listening to or reading has been generated with the help of AI.
When do you have to notify a user that they are using AI?
If you are using an AI chatbot, virtual assistant, or AI agent that communicates with humans, the user must be clearly informed that they are talking to an AI and not a human. The notification must be visible at first contact, for example with the message “You are talking to an AI system”.
This obligation applies to:
- chatbots for customer service,
- virtual assistants,
- AI agent,
- voice assistants,
- AI systems in e-commerce, finance, healthcare, and other industries.
What about AI-generated images, videos, and audio?
Companies that develop AI systems to generate images, video, audio, or text must enable such content to be technically recognizable as AI-generated. This can be achieved through metadata, digital watermarks, or similar technologies.
Attention should be paid to content that may look real, even though it is not.
Deep fake content must be labeled.
If a photo, video, or audio is generated or significantly altered by AI to make it look like a real person, event, or situation, the user must be clearly informed that it is AI content.
Examples include:
- a video in which a famous person appears in an event that never happened,
- AI-generated speech of politicians,
- synthetic influencers in marketing campaigns,
- AI-generated recommendations or testimonials from real people.
The label must be visible or audible to the end user. A technical designation alone is not enough.
What about AI-generated articles and posts?
If AI is used to create or significantly modify texts that inform the public about topics of public interest, such as politics, health, the environment or the economy, it is necessary to indicate that the content has been generated or processed by artificial intelligence.
However, there is an important exception. If the text has been reviewed by a responsible person or an editor who has checked the content and takes responsibility for publication, additional tagging is not required.
This text was initially generated by Copilot at the request of the author, and the author reviewed, refined and prepared the text for publication.
For most business blogs and professional articles, this means that human editorial control remains crucial.
What does this mean for marketing and business teams?
For most companies, the most important thing is to check:
- whether do they use AI chatbots on websites
- whether they publish AI-generated images or videos
- whether they use AI to create publicly published articles or press releases
- whether they have human content verification processes
- whether they educate employees on the use of artificial intelligence
The European Commission especially emphasizes the importance of AI literacy, i.e. understanding the possibilities and limitations of AI systems among employees and users.
What are the penalties?
Violations of transparency obligations can result in fines of up to €15 million or 3% of a company’s total annual global turnover, whichever is higher. A lighter penalty regime is foreseen for small and medium-sized enterprises.
Conclusion
The AI Act does not prohibit the use of artificial intelligence. On the contrary, it aims to increase user trust. The basic rule is simple: when AI interacts with humans or creates content that may look like it was created by a human, users should be clearly informed about it. Transparency is becoming one of the key requirements for the responsible use of artificial intelligence in business.
