A Context-Adaptive Safety Layer for AI-Based Academic Assistance

This paper presented CASL, a Context-Adaptive Safety Layer for AI-based academic assistance. The framework analyzes request context and risk before response generation, then selects among direct assistance, clarification, privacy protection, verification guidance, or safe refusal. Its design emphasizes minimal retention of sensitive information and evaluates both safety and educational utility. By combining context awareness, privacy protection, adaptive intervention, and measurable evaluation criteria, CASL provides a practical conceptual foundation for developing safer, reliable, and human-centred academic AI assistants

[1]   F. Miao and W. Holmes, Guidance for Generative AI in Education and Research, Paris, France: UNESCO, 2023. [2]   E. Tabassi, Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1, National Institute of Standards and Technology, Gaithersburg, MD, USA, 2023. [3]   C. Autio, R. Schwartz, J. Dunietz, S. Jain, M. Stanley, E. Tabassi, P. Hall, and K. Roberts, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile, NIST AI 600-1, 2024. [4]   OWASP Foundation, “OWASP Top 10 for Large Language Model Applications,” 2023. [5]   OWASP Gen AI Security Project, “LLM Top 10: 2025 Risks and Mitigations for LLMs and Gen AI Applications,” 2025. [6]   T. Thesen and S. H. Park, “A generative AI teaching assistant for personalized learning in medical education,” npj Digital Medicine, vol. 8, Art. no. 627, 2025.