CR4T: Rewrite-Based Guardrails for Adolescent LLM Safety

In three linesCR4T is a safety framework for adolescent-facing LLMs. Instead of refusing problematic requests, it rewrites unsafe responses into developmentally appropriate guidance. Combining lightweight risk detection with domain-conditioned rewriting, CR4T reduces unnecessary refusals while preserving benign intent.

Summary generated by Claude — human-verified