Beyond Detection: Redesigning Assessment and Governande of Generative AI at the Universidad Polit\'ecnica de Madrid (UPM)

The Shift from Detection to Redesign

The Universidad Politécnica de Madrid (UPM) has published a research paper that signals a potential inflection point in how higher education institutions approach generative AI. Rather than doubling down on detection tools or outright bans—the dominant "defensive" posture adopted by most universities globally—the UPM paper argues for a fundamental redesign of assessment and governance structures. This is not merely a policy tweak; it is a philosophical reorientation that treats GenAI as an embedded feature of the academic environment rather than a transient threat to be neutralized.

Why This Matters

The significance of this work lies in its timing and its institutional context. Most universities have responded to tools like ChatGPT with a predictable cycle: panic, followed by rushed adoption of AI-detection software, followed by student backlash and accusations of false positives. The UPM analysis correctly identifies this as a losing battle. Detection arms races are costly, unreliable, and pedagogically sterile—they train students to evade detection rather than to use AI responsibly.

By contrast, the UPM framework proposes redesigning assessments so that AI use is transparent, permissible, and even required in certain contexts. This aligns with emerging best practices in industry, where professionals are already expected to work alongside AI copilots. The paper’s emphasis on governance—not just classroom policy but institutional frameworks for ethical AI use—is particularly timely as European universities navigate the EU AI Act’s requirements for transparency and accountability.

Implications for AI Practitioners

For AI developers and product managers, this research carries three practical lessons. First, the market for AI-detection tools in education is likely to shrink as institutions shift from policing to integration. Startups building detection-only products should diversify into assessment design and governance consulting. Second, the UPM approach creates demand for AI systems that are transparent by design—tools that can document their own contributions to student work, enabling the kind of audit trails that redesigned assessments will require. Third, the governance frameworks proposed here may serve as templates for other regulated domains, such as professional certification or medical licensing, where the question of "what counts as original work" is equally fraught.

The paper’s most provocative implication is that the defensive posture may actually increase academic dishonesty by incentivizing students to hide AI use. If true, the UPM’s offensive redesign strategy is not just more ethical but more effective. For AI practitioners, the message is clear: build for integration, not detection.

Key Takeaways

• The UPM paper represents a shift from defensive AI policing to proactive redesign of assessment and governance structures in higher education.
• Detection-based approaches are increasingly seen as unsustainable due to cost, inaccuracy, and negative pedagogical incentives.
• The framework creates market opportunities for transparent, auditable AI tools that can be integrated into academic workflows.
• Governance models developed for universities may be applicable to other professional and regulatory domains facing similar AI integration challenges.