Who Is Behind the Content? How Can AI-Generated Content Be Identified?

Who Is Behind the Content? How Can AI-Generated Content Be Identified?

Artificial intelligence has become a natural part of corporate operations within just a few years. Marketing materials, reports, images, videos, and customer service responses are produced with AI assistance every day. As the technology advances, however, an increasingly important question emerges: do we still know whether a piece of content was created by a human or by artificial intelligence?

The European Union AI Act regulation seeks to address this challenge as well. One of the most important transparency requirements is the obligation to label AI-generated content, often referred to as the “watermarking obligation”.

Although it may sound like a technical detail at first, it is in fact one of the cornerstones of digital trust.

Why Is It Necessary to Identify AI-Generated Content?

Generative artificial intelligence has now reached a level where, in the case of a well-written text, a professional image, or even an audio recording, it is often no longer possible to clearly determine whether it was produced by a human or an algorithm.

This carries numerous risks: • misleading marketing communications; • deceptive or manipulated media content; • synthetic audio and video recordings; • blurred lines of authorship and accountability; • erosion of trust in digital content.

misleading marketing communication,

deceptive or manipulated media content,

false audio and video recordings,

the blurring of questions of authorship and responsibility,

the erosion of trust in digital content.

The aim of the AI Act is not to restrict the use of AI-generated content, but to ensure that users can make informed decisions about how they evaluate a given piece of content. The underlying principle of the regulation is simple: if artificial intelligence was involved in the creation of content, this must be recognisable.

What Does the AI Act Require?

Article 50 of the AI Act regulates the transparency obligations for general-purpose AI systems and generative AI solutions. The regulation requires that providers of AI systems apply technical solutions that enable the recognition and identification of AI-generated content.

The requirement is particularly important for widely used systems such as: ChatGPT, Claude, Gemini, Microsoft Copilot, Midjourney, and other generative AI solutions. There is a clear legislative intent behind the regulation: it must be made visible to users when a piece of content has not been created by a human author.

What Does the “Watermark” Mean in Practice?

In public discourse, this requirement is often referred to as a watermark; however, the regulation does not prescribe a single specific technological solution. The key is not the method itself, but the outcome: AI-generated content must be recognisable.

Visible Labels

The simplest solution is when the AI system directly indicates that the content was produced with the assistance of artificial intelligence.

For example: "This content was created with the assistance of artificial intelligence."

This approach is easy to understand, yet does not in itself provide protection against the removal of the label.

Digital Watermarks

A more advanced solution is digital watermarking, where information is embedded in the content that is not necessarily visible to the user but can be verified later with specialist tools. This approach is spreading particularly in the case of images, videos, and audio materials.

Metadata-Based Identification

Information about the origin of content can also be stored in the file's metadata. Such data may include, for example: the time of generation, the name of the AI system used, the model version, and the technical circumstances of production. This information can play an important role in a subsequent investigation or authenticity check.

The C2PA (Coalition for Content Provenance and Authenticity) standard is gaining increasing importance in establishing the provenance of AI-generated content.

C2PA is an open industry standard that supports the verification of the provenance and authenticity of digital content.

Behind the initiative stand global technology players such as Adobe, Microsoft, Google, Intel, and Sony. The aim of the standard is to enable the provenance and modification history of digital content to be tracked in a uniform manner. This is particularly important because, while the AI Act imposes obligations, industry standards are also needed for practical implementation. Based on current trends, it is clear that European regulation and international technology initiatives are pointing in the same direction: towards greater transparency of AI-generated content.

What Areas Does the Standard Not Resolve? What Questions Remain Open?

Although the regulatory objective is clear, practical implementation still holds many challenges. The following questions, for example, remain not entirely clear:

how identifying information is preserved across different platforms;

how texts, images, videos, and audio recordings can be handled in a uniform manner;

how resistant systems will be to manipulation attempts;

how compliance can be verified in a global digital environment.

The coming years are expected to be about not only the development of technology, but also the evolution of audit and compliance practices.

What Does Establishing the Provenance of Digital Content Mean for AI Developers and Providers?

Preparation for compliance should not be left until the last moment. Developing solutions that support the identifiability of content is expected to become a fundamental requirement of AI systems.

What Does Establishing the Provenance of Digital Content Mean for Companies Using AI?

AI compliance is increasingly less exclusively an IT question. Cooperation between legal, quality management, data protection, and compliance functions is becoming necessary.

For this reason, organisations should consider:

what AI systems they use;

what content is produced with AI assistance;

whether internal regulation or labelling practices need to be established;

how transparent communication can be ensured towards clients and partners.

What Does Establishing the Provenance of Digital Content Mean for Leaders Using AI?

The transparency of artificial intelligence use is increasingly becoming a reputational question as well. Organisations that communicate openly and responsibly about the use of AI can build greater trust among their clients and partners in the long term.

Deadlines and Preparation Regarding the Provenance Identification of Digital Content

The new transparency requirements of the AI Act are entering into force gradually; newly marketed general-purpose AI models must comply from 2 August 2025, while providers of models already placed on the market must comply by 2 August 2027 at the latest. Based on the agreement reached by the negotiators of the EU Parliament and the Council on 07 May 2026, and the EU Digital Regulation package submitted at the end of 2025 (OMNIBUS VII – Digital Omnibus), the 2 August 2027 deadline has been brought forward to 2 December 2026. Ensuring the identifiability of AI-generated content is not merely a technological question but also a regulatory, governance, and reputational task. Companies currently using generative AI solutions cannot wait until the deadlines: they must map their AI use, establish the necessary internal regulations, and satisfy themselves that their suppliers and AI service providers will also be able to meet the new requirements. Compliance does not begin on the day of the deadline, but with the first step of preparation. Timely preparation not only serves to reduce compliance risks, but also contributes to strengthening the organisation’s credibility and the trust placed in the use of AI.