The rapid adoption of generative AI has intensified debates about academic integrity, authorship, and assessment in higher education. However, discussions often focus on detecting AI use rather than examining a more fundamental question: what evidence of learning should students demonstrate in an environment where information and AI-assisted tools are widely accessible? Drawing on experiences as a Senior Lecturer and Academic Integrity Officer, this roundtable explores how academic integrity has become increasingly difficult to define and communicate consistently. Academics and students may perceive integrity as correct referencing, paraphrasing, AI disclosure, or compliance with assessment rules, while academics face growing challenges in determining the extent and appropriateness of AI use. These tensions raise important questions about the sustainability of traditional assessment approaches and misconduct-focused responses. The session invites participants to consider whether academic integrity should be reframed around demonstrating capability rather than detecting unauthorised assistance. Through facilitated discussion, participants will explore approaches such as authentic assessment, oral presentations, reflective defence, scaffolded tasks, intensive schools, collaborative problem-solving activities, and other integrity-resilient practices that provide credible evidence of student learning. The discussion aims to identify practical strategies for supporting staff and students while preserving academic standards in an increasingly AI-enabled educational environment.