The Commission published guidelines on the responsible use of generative AI in research, recommending avoiding its use in peer review and evaluation processes.
During the European Research and Innovation Days – a flagship event from the European Commission (EC) that took place on 20-21 March 2024 – EC Executive Vice-President Margrethe Vestager announced the publication of the living guidelines on the responsible use of generative Artificial Intelligence (AI) in research. These guidelines were prepared in close collaboration with the European Research Area (ERA) Forum, which gathers Member States representatives and stakeholder organisations representing key players such as universities, research and technology performing organisations, and research & innovation funding organisations.
To build this new set of recommendations, the drafting team based its work on pre-existing frameworks such as the European Code of Conduct for Research Integrity from All European Academies (allea) and the Ethics guidelines for trustworthy AI from the EC High-Level Expert Group on AI. Furthermore, the EC performed a thorough analysis of other existing guidelines (see pages 13-14 for the list of analysed guidelines). The ethical perspective from allea, in addition to the performed analysis, led to the four key principles for the new EU guidelines, which are (i) reliability, (ii) honesty, (iii) respect, and (iv) accountability.
In practice, the text provides key recommendations to researchers, research organisations and funding organisations and is meant to support the responsible integration of generative AI in research. Key points from the guidelines include – among others – that (i) researchers should avoid employing generative AI tools in critical tasks like peer reviews or evaluations, ensuring that their use of such AI respects privacy, confidentiality, and intellectual property rights; (ii) research organisations ought to promote the ethical application of generative AI and vigilantly oversee the development and utilisation of these tools within their institutions; and (iii) funding bodies are encouraged to assist applicants in adopting generative AI in a transparent manner.
This living document will be updated regularly to keep pace with this area’s very fast technological development. Its revision will be part of the preparatory activities on the new ERA action “AI in science”, which was proposed (and still needs to be adopted) for the next ERA Policy Agenda 2025-27. Regardless of the outcome of this ERA Action, the research community is encouraged to contribute their views and ideas on how to enhance future versions through the feedback form. The EC also made further information available on the ERA Platform and the AI in Science webpage.
Regarding the legislative background, the European Parliament adopted the AI Act on 13 March 2024, and the final text is currently being discussed in the Council and shall be approved shortly. Once approved by both co-legislators, the agreed text (Regulation) would then be published in the EU Official Journal and enter into force on the twentieth day following its publication. The measure will be fully applicable 24 months after its entry into force, except for (i) bans on prohibited practices (which will apply 6 months after its entry into force), (ii) codes of practice (9 months after entry into force), (iii) general-purpose AI rules including governance (12 months after entry into force), and (iv) obligations for high-risk systems (36 months after entry into force).