Main Content

Chat-AI

Please review the terms of use on the home page.

Marburg University offers all members of the university community (staff and students) the opportunity to use text-generating AI language models free of charge and in compliance with data protection regulations. This service is provided in collaboration with the AI Service Center for Sensitive and Critical Infrastructure (KISSKI) of the Society for Scientific Data Processing GmbH (GWDG) in Göttingen via the so-called Academic Cloud.

The web-based platform can be used to perform a wide variety of tasks, such as text generation, analysis, and translation. A range of language models (Large Language Models, LLM) are available, including open-source models that run on the high-performance computing (HPC) infrastructure in Göttingen. This ensures high data protection standards and strengthens the digital sovereignty of universities.

In addition, commercial language models, such as OpenAI ChatGPT, are available and are hosted in the Microsoft Azure cloud within European data centers. Usage is subject to a limited monthly quota, so availability may vary.

Frequently Asked Questions

What is the difference between open-source models and OpenAI (ChatGPT)?

Open-source models are freely available AI models that are developed transparently and can be operated by various institutions. They often run on local servers and are not subject to external data usage.
OpenAI (ChatGPT) is a commercial AI service whose models are developed and operated by OpenAI.
Does the AI learn from my input?

Open-source models: No. These models do not store or use input data for training purposes (see external service providers).
OpenAI (ChatGPT): OpenAI reserves the right to use your input data for training purposes.
Are my results or the outputs I generate saved or analyzed?

Open-source models: No. The generated content is neither stored nor analyzed (see external service provider).
OpenAI (ChatGPT): OpenAI may analyze generated content and use it for training purposes.
How is data protection ensured?

As a general rule, your input must not contain any personal data belonging to third parties!
Open-source models: The external provider is based in Germany and is subject to German and European data protection laws (see Legal Notice).
OpenAI (ChatGPT): OpenAI’s privacy policy applies. However, only your input is transmitted via the API used by GWDG—OpenAI does not receive your name, your university account, or your IP address.
Are there any costs associated with using it?

The external provider offers this service free of charge to all members of the University of Marburg; this applies in particular to the use of open-source language models.
Use of OpenAI (ChatGPT) is free for you; the HRZ covers the costs. There is a limited quota, which means availability may be restricted at short notice or may be exhausted for the current month.
Are there any rules for using it?

Yes, you can find them in the green box on the service's home page.
Is it okay to use generative AI for exams and term papers?

Exams and term papers are subject to the respective examination regulations and the principles of good academic practice. All external content must be clearly cited; failure to do so may be considered an attempt at academic dishonesty. Generative AI may be used solely as a supportive tool.
Who owns the copyright and exploitation rights to AI-generated content?

In Germany, copyright applies only to human creations. AI-generated content is therefore generally not protected by copyright. Exceptions may apply if:
- A direct copy or accidental resemblance to existing works is created.
- The user’s input reaches a sufficient level of creativity to allow for the assertion of their own copyrights.
Since the legal situation regarding AI-generated content has not yet been definitively clarified, there is no binding case law on the matter. This issue is currently the subject of intense debate.
The results are incorrect

The results generated by the service are not always accurate or precise and may contain inaccuracies, even if they appear detailed. Furthermore, the information provided may be incomplete or outdated, and the results may differ from our views. Therefore, you should never rely solely on the results without verifying their accuracy and correctness.