The paper analyzes state law risks associated with the use of generative language models (AI) in the education system. Its main focus is on the threat of distortion of the mechanisms through which students’ worldview is formed, thereby transforming a traditionally pedagogical task into an issue of national security. The authors argue that, despite its outward dialogic nature, AI lacks the capacity for reciprocal reflection and value-based transformation, which leads to the substitution of independent thinking with ready-made intellectual templates, the weakening of critical thinking, and a decline in motivation for the independent pursuit of knowledge. A particular legal problem is the cultural and value imbalance arising from the predominance of alien cultural codes in global models, which runs counter to the objectives of preserving national identity and may be regarded as an instrument of unlawful informational influence. At the same time, the authors acknowledge the constructive potential of these technologies for the humanization and personalization of education. As a legal conclusion, the article substantiates the need to move from the use of commercial models toward the creation of sovereign educational AI platforms operating within a special legal regime. The key tasks of the state are identified as the development of regulatory requirements for the algorithmic transparency, cultural adequacy, and pedagogical orientation of AI as a tool, as well as the legal regulation of its use for the protection of informational sovereignty and the constitutional rights of citizens in the educational sphere.