How to safeguard AI workloads with Unity AI Gateway Guardrails

This post introduces LLM Guardrails in Unity AI Gateway, a set of security controls for governing and securing AI workloads.

• •Built-in guardrails prevent PII leakage, jailbreak/prompt injection attacks, and unsafe content generation
• •Custom guardrails enforce specific business policies such as brand protection and competitive safeguards
• •Guardrails support multiple actions: sanitize input, block input, or block output based on business requirements
• •Log mode enables safe testing without disrupting live traffic, and evaluator endpoints can be tuned for performance
• •Inference tables provide detailed auditing and monitoring to track and refine guardrail behavior over time

This summary was automatically generated by AI based on the original article and may not be fully accurate.