Guardrail validation is the process of testing and verifying that an AI agents safety constraints function correctly before and during deployment.

Input filtering for safety refers to the systematic screening of user inputs before they reach an AI agent or language model, designed to detect and block malicious, harmful, or policy violating content.

Policy enforcement refers to the automated mechanisms that ensure AI agents, systems, and users operate within defined rules, limits, and regulatory boundaries.

Prebuilt policies are ready to deploy rule sets that govern how AI agents behave, make decisions, and interact with sensitive systems without requiring organizations to build governance frameworks from scratch.

Response format defines the structure and schema that an AI agent uses when returning data to applications, APIs, or downstream systems.

A safety engine is the protective layer within an AI agent system that monitors, validates, and constrains agent behavior in real time.

Input sanitization is the process of cleaning, validating, and transforming user provided data before an application processes it.

Input validation is the process of examining, filtering, and sanitizing all data that enters a software system before that data is processed or stored.

Prompt screening refers to the process of analyzing user inputs before they reach a large language model, or LLM, to detect and block harmful, manipulative, or policy violating requests.

A rule engine is a software system that executes predefined business logic by evaluating conditions and triggering corresponding actions.

Encryption access control combines cryptographic protection with permission systems to ensure that only authorized users and systems can decrypt sensitive data.

Merchant evidence collection is the systematic process of gathering, organizing and presenting documentation to dispute chargebacks and prove that a transaction was legitimate.