Generative artificial intelligence (AI) is reshaping undergraduate computer science (CS) education, and instructors, assessment de- signers, and computing education researchers need clear guidance on what that means for evaluation. Current systems emphasize au- tomated grading and functional correctness, leaving higher-order outcomes such as analytical reasoning, design justification, ethical reflection, and authorship verification underassessed. This study synthesizes literature on automated assessment, AI disruption, and assignment design and argues why a novel evaluation framework is needed for undergraduate CS evaluation in the United States. Evidence is organized into four claims: validity gaps in correctness- centric grading, equity concerns introduced by AI-mediated work, pedagogical limitations in assessing higher-order outcomes, and an integration gap between assignment frameworks and AI-aware evaluation. Together these claims justify a shift toward AI-resilient evaluation that preserves academic integrity and supports the de- velopment of knowledge, skills, and abilities required for success in research and professional CS roles. These skills include problem analysis, solution design, code quality, ethical judgment, commu- nication, and collaboration. This work is intended for instructors, assessment designers, and computing education researchers seeking practical, near-term adjustments that protect integrity and deepen learning in AI-rich classrooms.