Artificial Intelligence (AI) is reshaping science assessment by challenging traditional constructs, enabling new forms of measurement, and raising ethical considerations. Grounded in the K-12 Framework for Science Education, this chapter examines four major shifts in assessment accelerated by AI. First, assessment constructs must evolve to include human-AI collaboration, critical evaluation, and ethical reasoning within AI-mediated inquiry. Second, AI-resilient, performance-based assessments are essential for measuring complex competencies that AI cannot easily replicate. Leveraging AI’s ability to collect and analyze diverse data types, these assessments can reveal complex patterns of student learning over time to provide insights into students’ conceptual development. Third, while AI can automate the scoring of complex tasks, addressing validity, such as construct representation and fairness, still depends on human oversight and rigorous validation. We discuss trade-offs across AI scoring approaches and emphasize the need for construct-aligned rubrics and human oversight. Finally, we explore how AI enhances formative assessment practices in the classroom through real-time feedback, multimodal data analysis, and learner-driven engagement. Rather than replacing teachers, AI augments instructional decisions and helps capture the dynamic nature of science learning. Together, these shifts require a principled approach to AI integration that prioritizes validity and instructional value in next-generation science assessments.