Warren Beaumont. Implementing Probabilistic Safety Guardrails via Constrained Reinforcement Learning and Large Language Model Reasoning for Risk Aware Autonomous Decision Making. CIS 2026, 4.