EU Clarifies AI Transparency Law Exemptions

The European Union’s new AI transparency rules have introduced requirements for businesses and publishers to disclose when content has been created or altered using artificial intelligence. However, the legislation also includes several exemptions that may allow certain types of AI-generated or manipulated content to avoid mandatory labels.

The rules come from Article 50 of the EU Artificial Intelligence Act (AI Act), which focuses on transparency requirements for AI-generated content. The legislation requires certain AI-created images, videos, audio and text to be clearly identified so that users understand when they are viewing content produced or modified by artificial intelligence.

Although the law applies within the European Union, it can also affect organisations outside the EU if their AI-generated content is accessible to people located in member states.

However, Article 50 includes four important exemptions that may apply to some creators, publishers and AI system providers.

1. Artistic, Creative and Satirical Content Exemption

One exemption applies to AI-generated or manipulated content that is clearly artistic, creative, satirical, fictional or similar forms of work.

This mainly applies to deepfake images, videos or audio that are created as part of entertainment, creative projects or fictional productions. In these situations, transparency requirements are limited, meaning creators only need to make users aware that AI-generated or altered content exists rather than adding disruptive labels that affect how the work is experienced.

The legislation states that when AI-generated content forms part of an “evidently artistic, creative, satirical, fictional or analogous work”, disclosure should be provided in a way that does not interfere with the enjoyment or presentation of the content.

However, questions remain over how this exemption will be interpreted in practice. Terms such as “artistic” and “satirical” can be subjective, meaning different organisations may have different views on what qualifies.

2. Publisher Exemption for Human-Reviewed Content

One of the most relevant exemptions for publishers relates to AI-generated or modified text.

Under Article 50, organisations using AI to create or edit text that informs the public about matters of public interest are generally required to disclose that AI was involved.

However, the law provides an exception when the content has gone through a process of human review or editorial oversight, and a person or organisation maintains responsibility for the final publication.

This means publishers may not need to apply AI labels if human editors have reviewed the content and remain accountable for what is published.

The exemption could be particularly relevant for news organisations, websites and other publishers that use AI tools as part of their content production process.

However, the legislation does not clearly define what qualifies as sufficient human review. It remains unclear whether a quick review is enough or whether a more detailed editorial process will be required.

3. Standard Editing and AI Assistance Exemption

Another exemption applies to AI tools that are used for standard editing purposes or that do not significantly change the original content.

For example, AI systems used for minor improvements, such as correcting grammar, improving formatting or assisting with basic editing tasks, may fall outside the transparency requirements.

The exemption applies where AI does not substantially alter the original information or meaning provided by the user.

However, the term “substantially alter” is not clearly defined within Article 50, creating uncertainty around where the boundary lies between simple assistance and meaningful AI modification.

For example, an AI tool that rewrites a sentence may be considered basic editing, while completely transforming an article’s meaning or structure could potentially require disclosure.

4. Obvious AI Interaction Exemption

The final exemption applies to AI systems designed to interact directly with people.

Article 50 states that providers must inform users when they are communicating with an AI system, unless it is already obvious to a reasonably informed and careful person that they are interacting with artificial intelligence.

This means some AI chatbots or virtual assistants may not require additional disclosure if users can clearly recognise that they are not communicating with a human.

However, this creates another area of uncertainty. What makes an AI interaction “obvious”?

For example, is a chatbot clearly identifiable if it has a robotic voice or an AI-generated name? Or would companies still need to provide an explicit notice?

Remaining Questions Around Article 50

While the four exemptions provide some flexibility, several parts of the legislation remain open to interpretation.

The biggest challenge is that many key terms used in Article 50 are not precisely defined. Words such as “artistic”, “appropriate”, “human review”, “public interest” and “substantially alter” could lead to different interpretations across industries.

For publishers, one of the most important areas will be understanding how regulators determine whether AI-assisted content has received enough human oversight to qualify for an exemption.

Similarly, organisations using AI-generated images, videos or interactive tools will need to monitor future guidance to understand how the rules are enforced.

What This Means for Content Creators

The introduction of Article 50 represents another step towards greater transparency around artificial intelligence. As AI-generated content becomes increasingly common online, regulators are aiming to ensure that users can distinguish between human-created and machine-generated material.

Although the exemptions provide opportunities for some publishers and creators to avoid additional labelling requirements, organisations should carefully review how they use AI tools and whether their processes meet the conditions set out in the legislation.

As enforcement develops, clearer guidance will likely emerge on how these exemptions should be applied in real-world situations.