Generative artificial intelligence (AI) models are increasingly able to produce and combine sophisticated text, image, and audio. These advancements are challenging composers and teachers, as they work to reimagine and resist ways that composition and creative work are changing. This paper reports on one analysis in a larger study on multimodal composition with preservice teachers and other students interested in education. The authors explored ethical issues in multimodal composition, reporting on key themes across participant responses, composing sessions, and instructor/researcher reflections, including concerns and hopes for equitable access, bias and representation, cheating oneself and others, human creativity, and ideas for evaluating AI tools. The article attends to the concerns of future educators in designing recommendations for ways to incorporate (or resist) use of AI models in the classroom. Additionally, the authors propose a tool, the "Model Card for Education" template, and methods to guide evaluation of AI as part of the educational environment but also the broader systems of platform ecologies.