Perspective: Safety-Critical LLMs�From Guardrails to Runtime Policies
Abstract
We argue for runtime policy sandboxes, per-tool permissions, and auditable reason codes as mandatory for safety-critical LLM deployments.
Cite this article
Torres, E. (2024). Perspective: Safety-Critical LLMs�From Guardrails to Runtime Policies. Research Explorations in Global Knowledge & Technology (REGKT), 3 (10). Retrieved from https://regkt.com/article.php?id=283&slug=perspective-safety-critical-llms-guardrails-to-runtime-policies