2.1 Copyright

Who “owns” an AI-generated text? Legal clarity is essential, particularly in the higher education context.

The use of generative AI systems in higher education raises complex copyright issues. In particular, with AI models that generate text, images, music or code, it is unclear who owns the resulting content. In the European Union, no specific copyright regulations for AI-generated content are currently in force.

Under German copyright law (Section 7 UrhG), content generated purely by AI generally enjoys no protection, as it lacks human creative input. Copyright law requires a “personal intellectual creation”.

However, protection is conceivable if a human uses the AI as a tool and thereby makes a significant intellectual contribution of their own. This can be achieved through extensive editing, major alterations or the input of highly original, detailed prompts, such that the human contribution becomes decisive for the result. Individual words or brief sequences of words in prompts are generally not eligible for protection.

 

When can others use the AI output?!

If there is no human copyright in the AI-generated work, it is generally in the public domain and freely usable.

Important restrictions:

• Terms of use of AI providers: The terms and conditions of AI providers (e.g. OpenAI) specify how the output may be used, including commercially.
• Risk of copyright infringements through training data: AI systems are trained using copyright-protected material. The AI output could unintentionally contain protected content belonging to third parties. The responsibility for this lies with the user.
• Academic context (higher ed institutions): Higher education institutions generally require mandatory attribution for the use of AI tools in student work. Students must document the prompts and outputs used.

 

What higher education institutions must check before using AI tools

However, many AI systems themselves rest on a legally fragile foundation. They are based on extensive datasets that may contain copyright-protected works. This applies both to training data (e.g. from scientific publications, textbooks or image databases) and to user-generated content during system operation. Higher education institutions that use or develop AI systems must therefore carry out legal checks on several levels:

• Training data: Have copyright-protected works been processed without a licence? Are there usage rights for the database? When using Open Educational Resources (OER), care must be taken to ensure correct licence attribution (e.g. CC-BY).
• Generated content: Is this subject to copyright? If so, who holds the rights of use – the institution, the AI provider, or the users?
• Contract law: When using commercial AI services, the providers’ terms and conditions and licence terms must be reviewed – particularly regarding issues of rights transfer, liability and further use.

For higher education institutions, this means:

• When creating teaching materials using generative AI, lecturers should carefully check whether the content is based on protected sources or whether free reuse is permitted.
• In coursework and examination work, it must be made clear which text passages were written independently and which were created with the aid of AI. A corresponding labelling requirement can be enshrined in examination regulations or codes of conduct.
• When developing proprietary AI models or training models (e.g. for research purposes), it must be fully documented which data was used and on what legal basis this took place.

It should also be noted that not only German copyright law, but also international agreements (e.g. the Berne Convention, the TRIPS Agreement) or European directives (such as the DSM Directive 2019/790) may play a role. The latter contains, among other topics, special rules on so-called text and data mining (Section 44b UrhG), which are also relevant for research at universities. Accordingly, works may be automatically analysed for research purposes under certain conditions – but only if the rights holder has not expressly excluded this.

 

 

 

 

Recommendations for higher education institutions

• Teaching materials: Carefully check whether content is based on protected sources or freely available
• Coursework and examination work: Identify original and AI-generated text passages. Ensure that the obligation to identify such content is enshrined in examination regulations
• Own AI models: Document in full detail which data was used and on what legal basis this was done
• Observe international agreements and European directives: Take into account the DSM Directive 2019/790 and text and data mining (Section 44b of the German Copyright Act). Works may be automatically analysed for research purposes under certain conditions – the rights holder must not have expressly excluded this.
• Develop guidelines: Create your own guidelines for dealing with AI-generated content
• Training: Provide regular training for lecturers, researchers and students on copyright issues
• Involve IT, legal and media services: Work closely with the relevant departments
• Review platforms: Ensure that only legally compliant AI content is used on publicly accessible platforms

 

💡 Learning Summary Chapter 2.1: Copyright

• No copyright on purely AI-generated content: Content generated by AI without human creative input is not protected by copyright. Higher ed institutions should therefore ensure transparency regarding how content is created – particularly in teaching and examinations.
• Legally compliant use of training data and AI services: It must be checked whether AI models are based on copyright-protected works and whether clear licensing terms exist for commercial AI services. OER content must be correctly labelled (e.g. with CC-BY).
• Responsibility in teaching, research and development: Higher education institutions should establish their own guidelines on AI copyright, train students and staff, and comprehensively document the data and rights involved in the development and use of AI.