MercoPress. South Atlantic News Agency

Uruguayan faculty warns students score top grades with AI's help

Udelar authorities are reviewing the validity of these tests with highly unusual grades

Teachers at the University of the Republic's (Udelar) School of Psychology in Montevideo raised concerns about students using AI tools, like ChatGPT, in virtual tests, achieving perfect scores in just three minutes. In addition, over 75% of correct answers were recorded when compared to in-person exams.

Between 2023 and 2025, the number of students with excellent grades surged, correlating with the rise of generative AI use. Faculty reported these issues in a letter to the dean, demanding a review of virtual evaluations and potential cancellation of midterm exams. The dean acknowledged the issue's urgency, with the Faculty Council set to decide on actions.

Virtual tests, implemented during the pandemic, are now under scrutiny for continuation. Globally, AI use in education is common, with 89% of Spanish university students and 90% of UK students using AI for assessments. Uruguay’s Education Minister, José Carlos Mahía, noted AI’s pervasive challenge in education, urging technical safeguards to ensure genuine student work.

“I am not surprised because artificial intelligence is, let's say, challenging us every day. It was not a scenario that was not possible because today it challenges everything, everything, and education as well. And what should be done in these cases to avoid this? Well, you have to work with guarantees for the students, technical guarantees, and that what is done is indeed genuine, but it is challenging you because it has changed the axes, the truth,” he said.

Critics such as Uruguayan psychologist Silvana Giachero argue that the issue lies in outdated evaluation models, not AI, as tests vulnerable to AI suggest a lack of demand for critical thinking or creativity. “The paradox is that instead of teaching students to use these tools well, to critically evaluate their answers, to contrast sources, and to develop complex thinking, they demonize them as if the problem were the technology and not the obsolete model of evaluation that they continue to apply. Furthermore, if the midterms can be answered by an AI without specific training in the subject, the problem is not the AI: it is that these midterms do not require true critical thinking, nor application, nor creativity,” she pointed out on X.