Belgian AI scientists resist the use of AI in academia (apache.be)

A coalition of AI researchers led by Olivia Guest and Iris van Rooij has published an open letter and position paper calling on Dutch and Belgian universities to resist integrating commercial generative AI into teaching and campus software. The five demands—ban AI in classrooms, stop normalising hype, protect academic freedom, preserve critical thinking and avoid using AI in in-house systems—have drawn more than 1,400 signatories, including prominent figures such as Luc Steels, Katrien Beuls, Wim Vanderbauwhede and Nicky Dries. The move comes amid clashes at Ghent and KU Leuven over “embrace vs. ban” policies and signals growing academic pushback that could shape institutional rules on integrity, assessment and research practices. Technically, signatories argue that current large language models (LLMs) are statistical, associative “best-guess” machines that produce plausible but ungrounded outputs (so‑called hallucinations), obscure training data provenance, embed bias, and consume large energy and data resources—raising copyright and reproducibility concerns. They warn that reliance on LLMs undermines students’ development of reading, argumentation and verification skills, pollutes codebases, and may even reduce productivity (one METR study found a 19% slowdown for developers using AI tools). Rather than normalising prompt-driven workflows, the authors call for teaching foundational skills—how computation and models work, coding, source criticism—and urge university leadership to act to preserve the epistemic standards of scholarship.