The UK higher education regulator, the Office for Students (OfS), recently released a report stating that existing AI detection tools cannot reliably determine whether students used AI to complete their assignments. It recommends that universities re-examine traditional assessment methods such as take-home essays and open-book exams, and gradually move toward supervised assessments.
The report argues that AI detection tools lack sufficient accuracy and are prone to false positives. As generative AI models continue to improve, identifying AI-generated content is also becoming more difficult. Instead of relying on detection technology to track AI use, universities may need to adjust their assessment systems and reduce unsupervised, uncontrolled forms of assessment.
At present, more and more universities are beginning to adopt in-person written exams, oral exams, and process-based assessment, evaluating student performance across multiple stages such as literature reviews and the research process, rather than scoring students solely based on the final paper. However, the report also notes that closed-book exams may be disconnected from real professional environments and may adversely affect some students with disabilities.
The report also warns that AI tools are changing how students learn and compete for grades. Many students believe that using AI is an effective way to earn high scores, and that not using AI may put them at a disadvantage. In addition, differences in capability between paid and free AI tools may also worsen inequality in educational resources.
This trend has attracted attention across the global education sector. New York City and Norway have previously introduced restrictions on students’ use of generative AI. At the same time, OpenAI launched ChatGPT for Teens and Study Mode, attempting to guide students to solve problems independently, but these features still face the risk of being bypassed.
As AI becomes increasingly integrated into education, the core challenge for universities is shifting from identifying AI cheating to rebuilding learning assessment systems, seeking a new balance among technology application, academic integrity, and educational equity.
