Manage AI Guardrails Data Classifications

AI guardrails data classifications are designed for use in AI Guardrails Rules , which protect against data loss or security violations in prompts and responses exchanged between users and AI applications. AI guardrails data classifications serve the same purpose as data classifications used in Real-Time and SaaS API rules—they describe the type of data the Data Loss Prevention policy will monitor or block.

AI guardrails rules and data classifications help ensure that AI applications used in your environment adhere to policies and standards you have established. AI guardrails protect your privacy and security, ensure regulatory compliance, and improve customer trust. AI guardrails data classifications focus on three areas of concern to protect your systems:

Security guardrails protect against prompt injection attacks, ensuring that prompts are safe for the AI model and will not cause the model to perform inappropriate activities such as deleting critical data, exposing sensitive data, or granting unauthorized access to resources in your environment.
Safety guardrails prevent the AI application from disseminating inaccurate information or information that could be used in a harmful manner, such as instructions for building a bomb or overriding safety precautions in heavy machinery.
Privacy guardrails prevent the AI application from disseminating personal Information Identifiers (PII) that may identify an individual, such as financial account numbers, medical records, passport or government identification numbers, or credit card numbers.

Like data classifications used in Real-Time and SaaS API rules, you can use built-in AI guardrails data classifications provided by Secure Access in your AI guardrails rules, or create and use your own customized AI guardrails data classifications.

Like data classifications use in Real-Time and SaaS API rules, AI guardrails data classifications are groups of data identifiers combined for the purpose of monitoring or blocking closely related content. Data identifiers describe the content the Data Loss Prevention policy monitors or blocks. Secure Access provides a set of built-in AI guardrails data identifiers; you cannot create custom AI guardrails data identifiers.

Individual Data Identifiers < Manage AI Guardrails Data Classifications > Create an AI Guardrails Data Classification