Like the EU’s General Data Protection Regulation (GDPR) in 2018, the EU Artificial Intelligence Act could become a global standard, regarding safeguards, limitations and rights on the use of Artificial Intelligence.. In February 2024, the European Parliament endorsed the provisional agreement on the AI Act that ensures safety and complies with fundamental rights. Read the press release here.

As AI is being increasingly used in research projects, this new regulation might have implications for you as a researcher.

Policy

UU is working on general policy on the use of AI, but it is unclear when that policy will be in place. Until then, the Faculty of Science applies the following guidelines.

• The use of AI must meet the requirements of UU’s Security Control Framework, which states: “AI tools do not use production data, including user prompts, to (further) train the underlying AI models. This also prohibits the entry of data (including questions from which sensitive matters can be derived) with basic confidentiality or higher in tools where it is not explicit that this data may not be used. AI tools are only allowed if they assure the UU that all UU prompts and answers are not used to train the model. Models trained on UU training data may only be used within the context of the UU.” (Security measure IS.13.007)
• Furthermore, the SCF states that sensitive data should only reside in organizationally managed environments. Before sensitive data leaves the organization, official relations including agreements with the receiving party about data storage and handling exist. (Security measure IS.8.009)
• The use of AI must comply with existing laws and regulations such as the GDPR.

How to proceed when you want to use an AI tool

If you want to use AI in (research) projects, the way to proceed is:

• Always contact the faculty Data Steward Frans Huigen.
• Have us help you out to:
  • Carry out a data classification so that it becomes clear whether the use of AI is justified.
  • If you are working with personal data, also perform a privacy scan.
  • Document the choices made in the project to ensure secure use of AI.

Important things to consider

• If your use of an AI tool changes (for example, because students you supervise are going to use it), you have to reconsider whether you are still staying within the rules and guidelines. In that case, follow the steps in the previous section.
• It is possible that in the future there will be a UU policy for the use of AI, which explicitly states, or from which implicitly follows, that the use of a specific AI tool is no longer allowed at UU. In that case, you will have to comply with this, so make sure you have an exit strategy for yourself.
• Research funders can have a policy for the use of AI within projects they fund. In other words, there may be no objection to you using an AI tool as far as UU is concerned, but that limitation can certainly still apply to your project budget. You will have to check this yourself with your research fund (if there is one).
• Restrictions may apply when you want to publish results. Some publishers or journals will not allow you to have used AI in your research, or will impose additional conditions on (the publication of) the research.
• Check whether your use of AI tools fits within making UU research FAIR.
• There are already several known cases of AI tools that have made up sources, quotes, or results (‘hallucinating’). So while an AI tool can help focus the thoughts and may give helpful insights, it is also your responsibility as a researcher to check the claims of an AI tool for their validity.