AI-based applications and content have become integral to our everyday lives, increasingly permeating the field of music education e.g. through algorithmically sorted listening recommendations, AI-generated lesson plans, or audio content on platforms like TikTok and YouTube. These tools rely on machine learning and deep learning, which, despite their statistical foundation, are deeply entangled with social structures. This paper explores the concept of a "musical habitus of the machine," examining how AI-based content inherits classifications, dispositions, and structures in music education. Three examples are analyzed: lesson planning with ChatGPT, image generation via Midjourney, and song production using Suno.ai. The findings highlight that the uncritical use of AI-generated content in music classrooms threatens the diversity of music education, potentially leading to a global homogenization of musical practices. The paper also suggests strategies to preserve diversity and agency in the face of these transformative processes.