There is little transparency from AI developers on what is done with personal data entered into the tool. The data is also often stored in cloud applications that are not always GDPR compliant, and storage is often outside the EU.  This raises issues for both personal data and new findings in scientific research. Feeding GenAI any of this data could be equivalent to disclosure, which may cause a data breach and/or prevent the filing of a patent for that discovery. For that reason, we ask you not to enter confidential information, such as personal or research data, into GenAI applications. Cases used in lectures and seminars are also considered sensitive information and aren’t allowed to be entered into GenAI applications.   

If possible, request that the GenAI tool does not train its algorithm with your data. You can request ChatGPT to not use your entered data for training the model via 'Make a privacy request'. Note that this does not relieve you from the previous guideline not to enter privacy-sensitive, IP-protected or copyrighted material into GenAI.

GenAI tools lack transparency about the sources used when generating output, increasing the likelihood of plagiarism. The risk of copyright infringement is an additional risk. The databases used by GenAI tools contain source material that leaves unclear whether original authors gave consent for usage of that material, or whether any copyright was respected. Be sure to paraphrase and reference properly (see above section on Citing GenAI APA-style) and apply thorough source research (see Verify GenAI below). Do not enter any source material written by a third party, such as an article, email, information from lectures and seminars, text from lecture slides, or example cases, unless given explicit permission by the author. 

Use GenAI if permitted, but it is important to remain vigilant. It is not always possible to find out how the algorithms arrived at a particular conclusion and there is no transparency about what sources were used. GenAI’s purpose is to generate text that seems as plausible as possible, and veracity is not verified. In some cases, the LLM may fabricate answers entirely, this is referred to as the GenAI ‘hallucinating’. This means there is no guarantee that the system's output is actually correct. Verify results and answers, and look for existing source material to cross-reference any claims made by GenAI. You are considered responsible for any incorrect information in your work that may have originally come from GenAI. 

GenAI tools are often trained with large quantities of unknown data , which may be outdated or no longer representative of current standards. Developers or publishers give very little transparency about how or if a GenAI has any inherent biases, or options to mitigate that bias. This means any answers given by GenAI may perpetuate stereotypes or biases.    

Using GenAI effectively can sometimes be a skill. Asking an overly simple question can produce an overly general or vague GenAI output. Use ‘prompt engineering’ to increase the specificity and usability of the output. Avoid leading or loaded questions, or attempting to push the GenAI towards a specific conclusion. 

Know that the energy cost of servers for GenAI tools is very high, so use them only when they can add value.

GenAI applications can be a useful tool, but we expect students to be critical of the output at all times. The actual meaning is given by your own reasoning, critical analysis, creativity, and reflection.