In 2025, the rise of generative AI has forced higher education to confront a fundamental question: how should we assess student learning in a world where AI can write essays, generate code, and simulate critical thinking? Educational psychologist Jason M. Lodge issued a clarion call for universities to move beyond reactive strategies and instead re-imagine assessment from the ground up. His updated “scoreboard” of institutional responses—ignore, ban, invigilate, embrace, design around, and rethink—reveals that only the final option offers a sustainable path forward. Ignoring or banning generative AI is no longer viable. Tools like ChatGPT are now deeply embedded in student workflows, with one German survey showing a quarter of university students using them daily. Attempts to detect AI-generated work have proven unreliable, with tools easily fooled by simple text manipulations. Even advanced detection systems suffer from high false positive rates, particularly with technical or non-native writing. Meanwhile, the arms race of surveillance—lockdown browsers, AI proctoring, and keystroke monitoring—has created an atmosphere of distrust and inequity. Students from under-resourced institutions are disproportionately affected, while wealthier schools deploy expensive systems that only deepen the divide. Invigilation, too, is increasingly futile. New tools like Cluely and CheatingDaddy allow students to use generative AI covertly during exams by reading prompts and receiving real-time answers through hidden screens. These developments render traditional remote proctoring obsolete. The result is a system that penalizes honest students, rewards evasion, and fails to address the real issue: traditional assessments were never designed to measure authentic learning. The alternatives—embracing or designing around generative AI—offer interim solutions but fall short of systemic change. Some institutions, like Ohio State University and many in China, now mandate AI use to build digital fluency. This approach encourages students to treat AI as a collaborator rather than a threat. Assignments that require students to critique, refine, or build on AI-generated content foster AI literacy and critical thinking. However, without clear guidance, students may still use AI to bypass deep engagement, turning assignments into polished but shallow outputs. Designing around AI—such as returning to in-person exams, oral defenses, or project-based learning—can be effective. Oral assessments, in particular, have proven resilient. Jason Lodge’s implementation of interactive orals with 230 postgraduates at the University of Queensland revealed deeper understanding than written exams ever could. Students could not fake insight in real-time dialogue, and misconceptions were identified immediately. Similarly, project-based assessments that require process documentation, peer feedback, and iterative development make AI-generated shortcuts ineffective. Yet these approaches are labor-intensive and not scalable across large classes. They also risk excluding students with disabilities or those in remote locations. While valuable, they are not a complete solution. True reform lies in rethinking assessment’s purpose. Teachers must ask: what do we truly want students to learn? If the goal is rote memorization or formulaic writing, then AI will always outperform. But if the goal is critical thinking, ethical judgment, creativity, and adaptability—qualities AI cannot replicate—then assessment must reflect that. This means shifting from efficiency to validity. Human judgment remains essential for evaluating nuance, growth, and originality. Automated systems may score consistency, but they miss the spark of insight. Assessments should allow multiple modes of expression—writing, speaking, visual media—and emphasize process over product. Students should submit drafts, reflections, and peer reviews to demonstrate authentic engagement. Furthermore, assessment must integrate AI literacy. Students should be taught not just how to use AI, but how to interrogate it—detecting bias, verifying claims, and understanding its limitations. Assignments that ask students to compare their work with AI-generated content, or to challenge an AI’s argument, turn AI into a sparring partner rather than a shortcut. Finally, assessment must cultivate intrinsic motivation. When students see assignments as meaningful and connected to their interests, they are less likely to cheat. Choice-based projects, peer collaboration, and creative formats like podcasts or zines increase ownership and reduce reliance on AI. Ultimately, integrity in the AI age is not about preventing cheating, but about ensuring authenticity in learning. The future of assessment lies in human-centered design—where technology supports, rather than replaces, the educator’s role in fostering deep, thoughtful, and ethical engagement with knowledge.