Recommendation on the Use of Generative AI Tools at the Friedrich Schiller University Jena

By “generative AI” we mean computer models capable of autonomously generating new content—such as images, texts, or music—based on human or machine input, after learning patterns and structures from a training dataset.

AI-based tools have immense potential to increase the efficiency and effectiveness of our work. Where generative AI can be used profitably and without risk, its use should be permitted. To guide and raise awareness for responsible, reflective, and legally compliant use, key points are compiled in the chapter "What must be considered when using generative AI tools?".

We must always bear in mind that although a text-generating AI may produce well-formulated texts based on statistical probabilities, it does not actually recognise their meaning or evaluate their coherence or factual accuracy.

To obtain commercial AI tools, please contact the URZ software service via the service portalExternal link.

The following procedure is established for the use of OpenAI products (ChatGPT Plus and API):

The costs are charged by collective booking to the PSP element specified in the service portal. The costs are averaged across all users in the respective billing period in order to distribute exchange rate fluctuations and foreign currency fees fairly. The URZ must receive the billing information from the OpenAI web portal in order to be able to carry out billing. The exact procedure differs depending on the type of product used.

ChatGPT Plus users will be charged a subscription of $20.00 per month plus transaction costs (exchange rate fluctuations and foreign currency fees of approximately $4.00). All users are obliged to download the billing documents provided by OpenAI in the web portal for their account on a monthly basis and make them available to the URZ (via the service portal in the same process in which access was requested).

When using the API, the costs are variable depending on the actual usage. An estimate of the expected monthly or annual costs is helpful for financial accounting. In each API access (organisation), several members can be authorised (‘team’). To give the URZ access to the billing information, softwarebeschaffung@uni-jena.de must be added as a user and authorised as an owner.

Data or information that is processed in the context of work-related activities should not be included in the training of an AI. This applies in particular to work protected by copyright, e.g. academic assignments by students, and personal data. Please check the settings in the AI tool before using it. There is usually an option to refuse use for training purposes. If this is not the case, we advise against using the tool in a university context. If possible, the input history and, if available, the memory function should also be deselected in the settings.

When entering data into an AI tool, please ensure that no data relating to a person or sensitive university information is entered or uploaded. This also applies to copyright-protected works and documents declared as confidential as well as information that may allow conclusions to be drawn about individuals via indirect channels. It is advisable to anonymise personal data before entering it (e.g. use a synonym instead of a name).

When entering your own personal details in AI tools, it should also be borne in mind that user profiles with personal information may be created and made available to third parties.

Generative AI tools should not be used as the basis or even the final decision for personal decisions (see Art. 22 General Data Protection Regulation). For example, AI should not be used for the selection of applicants, the assessment of students' examination results or third-party funding applications by researchers. The processing of data in an AI on the purposes of which certain personal aspects are ‘evaluated’ is not permitted.

Please always bear in mind that AI does not recognise the meaning, significance or implications of the outcome it creates.

The result of generative AI should be treated with caution and evaluated carefully. The information generated by the system may be incorrect and correlations may not be represented correctly (phenomenon of confabulation). Users should also check the accuracy of references. Discrimination and inequality in the AI-generated work should also be checked so that certain groups are not disadvantaged and existing social injustices are not reinforced, as AI can create or reinforce distortions or bias. Users are ultimately responsible for critically scrutinising the generated content and ensuring that it is correct and ethically acceptable.

It is also important to ensure that protected works or parts thereof are not reproduced. The rights of the author continue to exist, even if the AI has generated the result.

Be open about the use of generative AI. Its use should be clearly labelled and its use and underlying purposes openly communicated. It is also advisable to keep documentation on the use, for example when working on larger texts. In addition, the AI tool should also be named as the source for AI-generated images. For example, in the multimedia area of our content management system for image documents, the value ‘AI-generated’ can also be selected in the ‘Author’ field.