As of 2 August 2026, businesses must comply with new transparency requirements for AI-generated content. Artificial intelligence has become an everyday tool: it is used to write social media posts, generate images and videos, create advertising campaigns and even synthesize voices. But this raises an important question – does the audience have the right to know when content has been created with the help of AI?
From 2 August 2026, the European Union begins applying the transparency obligations set out in Article 50 of the EU AI Act. These requirements impose specific obligations on both AI system providers and organisations or individuals using AI systems in a professional context. Continue reading to discover what marketing and communications professionals should know and consider.
1. Does all AI-generated content have to be disclosed?
The AI Act does not require every text drafted with ChatGPT or every AI-generated image to be labelled “Created with AI”. Instead, it defines specific situations in which organisations using AI systems must inform the public about AI involvement.
For social media and communications professionals, two categories are particularly relevant: deepfakes and AI-generated or AI-manipulated text concerning matters of public interest when published to inform the public and not covered by the exemption for human review or editorial oversight.
In practice, this means businesses do not need to label every Instagram post simply because AI was involved in its creation. The key question is not “Was AI used?” but rather “Does this specific content fall under the transparency requirements of Article 50?”
2. The greatest attention should be paid to content that looks real
AI makes it possible to create highly realistic images, voices, and videos of people. However, this is precisely one of the most significant regulatory risks. For example, a company could create a video featuring its CEO, a public figure, or an event that never actually took place. If such content appears authentic, it could be deepfakes, which must be indicated when publishing it.
This must be taken into account when creating AI-generated images and videos of people, artificially generated voices, facial or voice transformations, artificial versions of real events, and virtual people who resemble real individuals. This means that not every AI-generated human image, voice, or virtual avatar is automatically deepfakes – It is necessary to assess whether the content could give the impression that it is authentic.
The European Commission has also created special icons that can be used to identify AI-generated content. While their use is not the only possible solution, the labeling must be clear and easily understandable to the public.
3. Human review becomes increasingly important
Another key requirement for communications specialists is AI-generated text on issues of public interest. If such content is subject to human review or editorial control prior to publication (i.e., it is proofread, edited, adapted, verified, etc.) and a specific person or company is responsible for its publication, the labeling requirement set forth in Article 50(4) does not apply to such text.
This does not mean that a mere “glance” at the text generated by AI is sufficient. There must be genuine human involvement. AI can help create content, but a human must be responsible for what is published. That is why it is worthwhile for companies to establish clear guidelines—content generated by AI must be reviewed by a human before publication.
4. Responsibility does not rest solely with the AI provider
A common misconception is the belief that “We’re just using ChatGPT, so its developer is responsible for compliance.” The AI Act distinguishes between the provider and the user of an AI system. A company that uses an AI system in its professional activities is, in certain cases, itself a regulated entity—that is, a system user.
For example, an AI system provider has certain obligations—including those related to the technical, or machine-readable, labeling of content generated by the AI. In turn, a company that uses this system may be required to inform people about deepfakes or text generated by AI. Therefore, a company needs to know not only which AI tool it is using, but also for what purpose it is using it.
5. Failure to comply with the requirements of Article 50 may result in significant penalties
The MI Act is not merely a set of recommendations. Member States must establish effective and proportionate sanctions for violations of the MI Act’s requirements. For failure to comply with the transparency obligations set forth in Article 50, the maximum administrative fine is up to 15 million euros or 3% of the company’s total worldwide annual turnover in the preceding fiscal — whichever is higher. For small and medium-sized enterprises, including startups, the maximum applicable threshold is the lower of the relevant monetary amount or percentage.
At the same time, this does not mean that a company automatically faces a penalty for every incorrectly labeled social media post. Penalties are determined based on the circumstances of the specific violation, including its nature, severity, duration, the company’s degree of liability, and other factors.
Technology will continue to advance faster than regulation
Alexander Morad, a digital marketing and AI expert, entrepreneur, speaker, and author of the international bestseller “AI OR DIE,” notes: “The rise of AI is inevitable, but regulation alone will not be enough to control AI, as technology is advancing much faster than legislation. The greatest concern is the risk that the European Union will regulate faster than it can implement innovations, and thus overly complex or costly compliance could harm EU startups, while large companies—which already have legal teams and the budget to ensure compliance—will be the ones to benefit.”.
For marketing teams, AI proficiency is far more important than bureaucracy—a ChatGPT subscription alone isn’t enough; proper employee training is also necessary. AI is increasingly becoming infrastructure rather than just a tool, so the question is no longer whether to use it, but how. In my opinion, the AI Act isn’t a finish line, but rather a safety barrier that may slightly slow down the technology’s development. However, technology will continue to evolve, and that means regulation will always strive to keep up with it.”
What Should Businesses Do Right Now?
The new regulation does not mean organisations should stop using AI. However, there are several practical steps businesses can take today:
- Identify where AI is being used across the organisation.
- Ensure meaningful human review before publishing content.
- Pay close attention to potential deepfakes content.
- Establish clear internal guidelines for AI use.
AI Content: Less Uncertainty, More Transparency
The new requirements do not mean businesses must add “Created with AI” to every social media post. The purpose of the AI Act is to ensure that people can recognise content that may mislead them about its origin or authenticity. That is why the primary focus is on deepfakes certain publications concerning matters of public interest and other situations where transparency is particularly important.
For businesses, this means adopting new principles in communication. The use of AI is not a problem in and of itself. The problem arises when a person cannot understand what they are actually seeing or with whom they are communicating. That is why now is the time to take transparency, reliability, and human accountability for published content much more seriously when using AI.
Staying up to date on the latest information and changes in the industry is especially important. So don’t miss this opportunity—on October 8, 2026, MI expert Alexander Morad will lead a masterclass in Riga titled “Power of AI Skills: Become a Future Marketing Leader.” During the masterclass, you’ll learn how to consciously follow trends, try out new approaches, and shape your professional future alongside AI—all with focused content and no fluff. Learn more about the event here!
Sources of information used in the preparation of this article:
EN 
LV