Go back

EU publishes guidelines for using generative AI in research

Image: EU Science & Innovation [CC BY-ND 2.0], via Flickr

RI Days: Commission executive vice-president says artificial intelligence has “vast” potential for research acceleration

The EU has published guidelines for the responsible use of generative artificial intelligence tools in research, aiming to ensure the adoption of a coherent approach across Europe.

Generative AI like ChatGPT, which are able to produce content such as text and video based on prompts and prior training, have shot to prominence over the past year due to their increased abilities resulting from faster computing power.

“AI is transforming research, making scientific work more efficient and accelerating discovery. While generative AI tools offer speed and convenience in producing text, images and code, researchers must also be mindful of the technology’s limitations, including plagiarism, revealing sensitive information or inherent biases in the models,” said the European Commission, which helped to produce the guidelines.

Published on 20 March, the guidelines provide “simple and actionable recommendations” for researchers, institutions and funders. They were devised in collaboration with countries and stakeholders involved in the European Research Area initiative for raising research standards.

They are based on four principles: that research must be of reliable quality; that honesty is needed in research and its reporting; that researchers must respect their colleagues, research participants and society more broadly; and that researchers must take accountability for their actions.

Critical approach and transparency needed

Example guidance includes that researchers should maintain a critical approach to using the outputs of generative AI, being aware that such tools can be biased and can present falsehoods as fact.

Researchers should also be transparent about their use of generative AI, and should not upload sensitive unpublished work to AI tools, the guidelines say.

Research organisations should train researchers in using generative AI, they say, and funders should be receptive to the ethical use of such tools.

EU research commissioner Iliana Ivanova called on the research community to “join us in turning these guidelines into the reference for European research”.

The Commission said that, as generative AI is evolving, so too will the guidelines be updated in response to feedback.

Vast potential, but need for education

Speaking at the Commission’s Research and Innovation Days event in Brussels on the day the guidelines were released, the EU institution’s executive vice-president for digitisation, Margrethe Vestager, said AI has “vast potential to accelerate science”.

“I think it’s true for every scientific field, from medicine to anthropology to linguistics,” she said. But she added that a “nuanced approach” to its use is needed to make researchers “comfortable” with using it in their work, and that the guidelines should help with this.

Another speaker at the event—Emanuela Girardi, who founded the Pop AI association to investigate the impacts of AI on society and is president of the AI, Data and Robotics Association public-private partnership—echoed the guidelines in saying there is a need to educate “every” researcher in using AI.

She added: “We really need to make sure that all scientists know how to use this tool and can use it in a responsible way.”

Additional reporting by Craig Nicholson in Brussels