The widespread adoption of Generative AI (GAI) since 2022 has fundamentally challenged traditional assessment methods in engineering education, exposing the weakness in assuming submitted artefacts reliably represent student thinking. This article's purpose is to provide concrete examples of how GAI has changed formative and summative assessment in engineering courses. Through comparative case studies across five CDIO institutions in Norway, Denmark, Finland, France, and Singapore, we examine GAI implementations involving over 400 students. All cases are characterized using the Traffic Light Model by Arene (2024), which categorizes AI use as required (blue), prohibited (red), reported (yellow), or allowed (green). Our findings reveal that effective GAI integration requires thoughtful assessment redesign—structured implementations with clear pedagogical frameworks demonstrated positive outcomes, whilst less structured approaches can lead to student over-reliance and reduced learning. The evidence indicates thoughtful assessment redesign grounded in constructive alignment is required, where AI use decisions are tied explicitly to learning outcomes. Effective integration demands shifting from product-based to process-based assessment methods, emphasizing evaluation of student reasoning, judgement, and capacity to work responsibly with AI.
