UPM paper argues AI detection is dead end, proposes six-dimension adoption framework

A new paper from Universidad Politécnica de Madrid (UPM) argues that universities are wasting resources on GenAI detection, which is an unsustainable arms race. The authors—Jessica Díaz, Sonia Linio, Fernando Pescador, and Daniel Martin-Fabiani—contend that generated text is increasingly hard to distinguish from human writing, and detection tools still misfire too often to be trusted. They side with pioneering universities that pursue deeper adoption, assuming any policy based on prevention or sanction will fail. The paper calls for a coordinated shift: clear course-by-course rules for GenAI use, redesigned assessment toward authentic, interdisciplinary tasks that foster critical thinking and learner autonomy, and a serious AI-literacy program that treats students as critical co-creators rather than passive users.

Beyond pedagogy, the challenge spans organizational, technical, operational, legal, and economic dimensions. UPM is developing a strategic and sustainable AI policy structured around six dimensions, where AI acts as an enabler of student autonomy and pedagogical innovation, not a threat to be policed. The framework aims to balance innovation with governance, addressing everything from infrastructure and training to ethical guidelines and cost models. This positions UPM as a case study for institutions seeking to move beyond detection and embrace GenAI as a transformative educational tool.