Generative artificial intelligence tools have entered university writing instruction faster than institutional policy has been able to respond, and the ways in which students use them are shaped by cultural and institutional context as well as by the technology itself. This paper compares how undergraduate students in Turkey and Egypt use generative AI tools in academic writing and how instructors adapt assessment in response. The study employed a sequential mixed-methods design combining a questionnaire survey of 480 undergraduate students, 246 at a Turkish university and 234 at an Egyptian university, with eight focus groups involving 52 of the surveyed students and interviews with 16 writing instructors. Survey and focus-group data from 480 students reveal distinct usage norms in the two settings. Turkish students reported more frequent use for drafting and structural planning and were more likely to disclose use to instructors, whereas Egyptian students reported greater use for language correction and translation and were more cautious about disclosure, reflecting differences in institutional signals and perceived sanctions. Despite these differences, students in both countries expressed a shared demand for transparent AI-use policies that specify permitted and prohibited uses at the level of individual assignments. Instructors in both settings had shifted towards process-based, in-class and oral assessment but reported limited institutional guidance. The paper argues that assessment adaptation and policy transparency are complementary and offers recommendations for institutions in comparable contexts seeking to integrate generative AI into writing instruction without undermining learning or academic integrity.