Updated EU research guidance addresses AI agents

On 16 October, the European Commission updated its guidance on the responsible use of generative artificial intelligence in research. The new version also addresses AI agents capable of more autonomous task execution. It can help Hungarian universities and their research partners design institutional workflows.

The full document stresses human responsibility in peer review and the evaluation of funding proposals. Data processed and conclusions drawn by agents require validation by a researcher. The publication offers recommendations; it is not a newly adopted research-funding regulation.

As our own practical consideration, a Hungarian research organisation should establish in advance which task it delegates to the tool, who reads back the result and how the work is documented. Even a fluently written summary may need separate source verification. Technical performance and the credibility of a research result are therefore separate assessment questions.

Our earlier report on the EU’s AI incident assessment describes the risk background, while the new sectoral AI-use figures examine adoption. The guidance is a living document with an open feedback process; the announcement gives no date for another revision. The next practical question is how institutional rules will be aligned with actual research workflows.