GOVERNANCE-CONSTRAINED ARTIFICIAL INTELLIGENCE IN COMPLIANCE-CRITICAL TRAFFIC SAFETY EDUCATION A REGULATION-BOUNDED INSTRUCTIONAL ARCHITECTURE FOR CERTIFIED PUBLIC SAFETY SYSTEMS

Abstract

Traffic safety education operates within legally regulated environments where instructional precision, statutory fidelity, and public trust are foundational requirements. Although artificial intelligence has demonstrated significant potential to improve educational accessibility and learner comprehension, its deployment in compliance-critical domains remains constrained by risks of hallucination, regulatory drift, and legal liability exposure. This study formalizes the Regulation-Bounded Artificial Intelligence (RBAI) Model, a governance-constrained instructional architecture designed for integration within state-certified traffic safety education systems. Developed during an active regulatory certification process in California (2024–2025), the RBAI Model embeds regulatory authority directly into system architecture through retrieval constraints, structured governance validation, and full audit traceability. Comparative compliance analysis demonstrates that governance-bounded AI architectures significantly reduce instructional variance and eliminate hallucination pathways relative to open generative systems. The findings establish a replicable governance framework for responsible AI deployment in public safety education and other compliance-sensitive domains.