Policy Enforcement
The process of applying defined rules and policies to control, restrict, or permit AI system inputs, outputs, and actions.
Explore more about Guardrails
Related terms
A software component that executes predefined logical rules to validate, filter, or control AI inputs and outputs.
Input ValidationThe process of verifying and sanitizing user inputs or external data before they are processed by an AI system to ensure they conform to expected formats, constraints, and security policies, helping prevent errors, misuse, prompt injection, and malicious inputs.
Output ValidationThe process of checking AI-generated outputs against expected formats, schemas, rules, or safety requirements before they are accepted or used.
Content FilteringThe process of detecting, blocking, or modifying content that violates safety policies, compliance requirements, or application rules.
Safety FilterA mechanism designed to detect and block toxic, harmful, or policy-violating content in model inputs and outputs.
ModerationThe process of detecting, classifying, and handling harmful, unsafe, or policy-violating content in AI inputs and outputs.