Over the past decade, artificial intelligence (AI) has become a familiar part of everyday life, but its role in scientific research has been evolving for much longer. Understanding how AI is used in contemporary science is increasingly important for shaping science education from kindergarten through college. This chapter examines how AI is currently integrated into different scientific disciplines, drawing on interviews with practicing scientists and reviews of disciplinary trends. Five key conclusions and recommendations about the roles of AI in science and science education are advanced: First, scientists work in collaborative networks (i.e., epistemic networks) to investigate the natural world. Complex machines and instruments play essential roles in knowledge production, and their unique epistemic features should be explored with students prior to AI integration. Second, AI plays different roles in distributed networks of scientific knowledge production (e.g., instrument, agent) that in turn have varying implications for epistemic dependence, trust, and vigilance. These roles should be differentiated, explicitly modeled, and scaffolded using epistemic heuristics for students. Third, AI-based scientific research shares similarities with traditional science practices, as well as notable differences in epistemic goals and scales. These differences may require re-evaluation of the constructs of science practices currently used in science education. Fourth, scientists employ AI across all science practices, indicating that widespread integration of AI into educational materials will be necessary. Finally, scientists learn about AI in fragmented and idiosyncratic ways in formal and informal contexts, and standardization of AI norms and protocols remains limited. How students learn about appropriate use of AI in science education will require a more principled, structured, and effective strategy.