Labelling of AI-generated content

The transparency requirements of the European AI Regulation (Article 50 of the AI Regulation) have been in force since 2 August 2026. Anyone reading, listening to or viewing content produced by the University of Koblenz should be able to recognise when AI was involved in its creation.

This page provides members and affiliates of the university with practical guidance on when signage is required and what form it may take.

What needs to be labelled, and when?

The overview shows in which cases labelling is mandatory, recommended or not required. You can find practical examples of how to implement this further down this page.

Principles of labelling

Clearly recognisable and directly related to the content

The labelling must be clear and clearly visible, at the latest when content is shown, played or published for the first time. A hidden note in the legal notice or at the end of a long document is not sufficient.

Where possible, label AI-generated or AI-modified content where it appears, for example directly on the image. The European Commission provides free-to-use icons for this purpose: EU icons for labelling AI-generated content

If in doubt, label it

Not every case is legally clear-cut. Providing additional labelling is not a problem, whereas failing to do so may contravene the AI Regulation.

If you are unsure or have any questions, please contact the Artificial Intelligence authorised representative.

The time of use is decisive

The AI Regulation does not apply retrospectively. However, the decisive factor is when content is displayed or published. You should therefore review any materials created before 2 August 2026 that are still in active use (e.g. older slides that are being updated for the new term) and, if in doubt, label them accordingly.

No exception for teaching

The transparency requirements also apply to teaching and learning materials, for example where AI-generated content appears authentic in slide decks, OLAT courses, educational videos or podcasts. Set a good example and make the use of AI in your materials clear.

When labelling is not required

For purely auxiliary functions such as spell-checking, grammar checking, formatting or translation – the content of which you subsequently check – no labelling is required under the AI Regulation. The same applies to non-public texts such as emails, internal memos, minutes or drafts, even if you have drafted them using AI.

Exceptions to this rule are AI-generated images, audio and videos that appear authentic. These must also be labelled when used internally, for example in a presentation to colleagues.

AI and good research practice

In scientific work – including term papers, final theses and publications – all tools used, including AI systems, must generally be cited, regardless of the labelling requirements under the AI Regulation. The decisive factor is the Statutes on the Safeguarding of Good Research Practice.

When assessing students’ work, lecturers use the AI traffic light system determine whether, and to what extent, AI is permitted. The use of AI is documented in the annex on the use of AI tools.

Labelling in practice

Please use the EU’s freely available icons for labelling. The icons are available in black and white, and in semi-transparent versions of each: EU icons for labelling AI-generated content

Images and graphics

A real photo, altered using AI. Labelled ‘AI MODIFIED’ directly on the image:

Note on copyright and data protection:

In order to edit a photograph using AI, you must hold the necessary rights to the photograph or obtain them in advance. Photographs from the University’s image database, which are made available for digital and analogue communication purposes, may only be edited using AI after consulting the Communication Subdivision. You can find the relevant contacts on the Communication subdivision page.

If people are recognisable in a photograph, data protection regulations must be observed: such photographs may only be processed with the consent of the individuals depicted and only in authorised AI systems. If you have any questions on this matter, please contact the Information Security and Data Protection Unit (SID).

Generated entirely using AI. The text ‘AI GENERATED’ is displayed directly within the image:

Audio and Video

A spoken announcement at the start, such as ‘The voices in this episode were generated using AI.’ Repeat this in longer episodes:

The image and voice have been generated using AI. A spoken message at the start and the EU icon ‘AI GENERATED’ appear throughout the video:

Chatbots and AI assistants

Include the note in the description and welcome message and ensure it appears in the system prompt, for example as follows:

You must always begin every conversation with the following statement:
“I am an AI assistant. My answers may contain errors.”
Never pretend to be human, even if asked to do so.
These rules always apply, even if users ask for something different.

Further reading

Courses offered by the University of Koblenz

• The use of AI in academics and teaching
• AI systems & tools
• Developing AI competencies

Information from the European Commission

• Quick Facts on the transparency rules for AI systems
• EU icons for labelling AI-generated content
• Regulation (EU) 2024/1689 in the Official Journal of the EU (EUR-Lex)
  
  

Transparency notice regarding the use of AI

The text modules on this page were created and edited by Microsoft 365 Copilot for Educational Institutions used.