European AI ActHow the New Transparency Requirements for Artificial Intelligence Work
What applies to companies and other professional users?
The new transparency requirements stem from Article 50 of the European Union’s so-called AI Act. They are primarily aimed at companies, government agencies, media outlets, and other organizations that develop or use AI systems. The AI Regulation distinguishes between providers and deployers.
Providers are companies that develop and bring AI systems to market, such as OpenAI, Google, or Midjourney. They must ensure that their systems meet certain requirements. Since the beginning of August, this has also included the requirement that AI systems that generate content such as text, images, audio, or video files technically label this content in a way that makes it recognizable as artificially generated or manipulated. This labeling is primarily intended for other technical systems and platforms and does not necessarily have to be visible to users.
“Operators” are defined as those who use AI systems in the course of their own business activities. This includes, for example, companies that use an AI chatbot for customer service, media companies that publish AI-generated content, and organizations that use AI-generated images or videos for their communications. Depending on the area of application, specific disclosure requirements now apply to them.
Particular attention is being paid to so-called deepfakes. If AI is used to generate or manipulate realistic-looking images, videos, or audio files that are subsequently published, it must be disclosed that this content was artificially created or altered. However , works that are obviously fictional are generally not subject to this specific disclosure requirement. A fantasy video featuring fairies and dwarves, for example, is easily recognizable as artificial content and therefore does not need to be accompanied by an additional disclaimer. Special exemptions also apply to artistic, creative, satirical, or fictional works.
Context is also key when it comes to AI-generated text. If a text is published to inform the public about matters of public interest, it must generally be disclosed if the content was artificially generated or manipulated. This may apply, for example, to political content, news, or other public information. An exception applies if the content underwent a process of human review or editorial oversight prior to publication and a natural or legal person assumes editorial responsibility. In this case, the disclosure of the text’s AI origin may be omitted. A short story, poem, or recipe created using AI, on the other hand, is not subject to this specific transparency requirement, as these are generally not matters of public interest.
In addition to these regulations, there are other transparency requirements, such as those related to the use of emotion recognition or biometric categorization. However, these primarily concern specific applications and are less relevant for most companies in their day-to-day operations. An informative overview of all transparency requirements is provided by the European Commission on its website.
Key Points at a Glance
- The transparency requirements primarily apply to companies and professional users.
- Providers must technically label AI-generated content.
- Companies and professional users must disclose certain AI-generated content.
- This applies primarily to realistic deepfakes and texts on matters of public interest.
- Exceptions apply to fictional content and editorially reviewed texts.
What do the transparency requirements mean for regular users?
For people who use AI solely for personal purposes, the new transparency requirements will have few immediate changes. Anyone who uses ChatGPT at home, creates images with AI, or generates videos using AI does not automatically have to label this content themselves. The requirement to make AI-generated content technically identifiable applies to the providers of AI systems.
Private individuals are also not affected by the disclosure requirement. However, this applies only if they use AI exclusively for personal and not professional purposes. Within this private context, they may continue to create and distribute AI-generated content without disclosing it accordingly.
Nevertheless, users may become more aware of the effects of the new rules online in the future. Platforms and other digital services will use the technical labels provided by AI providers to provide information about the origin of content. As a result, it will be easier to tell—for example, on social media platforms—whether an image, video, or audio file was created using AI. However, this does not mean that all AI-generated content will automatically be marked with a visible label in the future. The type of content and the context in which it is used are decisive factors in this regard.
Key Points at a Glance
- Private users are generally not required to disclose their own AI-generated content.
- AI-generated content could become easier to identify online in the future.
- The new rules do not require all AI-generated content to be visibly labeled.
What can the new transparency requirements achieve?
The new transparency requirements will not prevent manipulated or misleading content from continuing to appear on the internet. Those who deliberately seek to spread disinformation will certainly not comply with labeling requirements. Furthermore, technical labels cannot prevent content from being copied, altered, or redistributed without the original information.
Nevertheless, the new rules serve an important purpose: they create greater transparency in a large part of everyday AI use. This is because a significant portion of the AI content generated today is produced via well-known and publicly accessible services and platforms such as ChatGPT or Meta AI. If these providers are required to incorporate technical labels, platforms and other digital services will be better able to identify AI-generated content and inform users about its artificial origin.
The transparency requirements are therefore aimed less at those who deliberately seek to deceive. Rather, they are intended to ensure that artificially generated content does not remain completely invisible in everyday digital life. This gives users better opportunities to assess the origin of information, even though complete protection against deception and manipulation on the internet will still not be possible.
This is precisely why media literacy remains a central component in dealing with AI-generated content. While transparency requirements and technical labels can help users better assess the origin of content, they do not replace the ability to critically evaluate information. Children and adolescents, in particular, must continue to learn how to question sources, contextualize information, verify facts, and distinguish reputable sources from unreliable ones in order to navigate the internet safely and confidently.
