How to configure guardrails to enhance privacy protection and security

This guide covers how to configure guardrails (automatically detect and redact PII, such as names, email addresses, phone numbers, and social security numbers) on the Maybe* platform to enhance privacy protection and security.

Guardrails on the Maybe* platform are essential for ensuring privacy and maintaining the integrity of your AI Agents. These features help in detecting and redacting personally identifiable information PII and preventing jailbreak attempts that could compromise agent instructions and security controls.

Navigate to the Configure section on the Maybe* platform.
Select the Guardrails page to access the configuration settings.
Enable the option to automatically detect and redact PII, such as names, email addresses, phone numbers, and social security numbers.
Note: Jailbreak Detection is always activate. This feature prevents any attempts to bypass agent instructions and security controls, keeping your AI Agents focused on their tasks.

Troubleshooting

Issue	Possible Cause	Suggested Solution
PII not being redacted	Guardrails not enabled	Check the Guardrails page to ensure detection is on
Jailbreak attempts undetected	Jailbreak Detection not activated	Verify that the Jailbreak Detection feature is enabled
Errors in configuration	Incorrect settings in the Guardrails configuration	Review and correct settings in the Configure section

In summary, configuring guardrails is crucial for protecting sensitive information and maintaining the security of your AI Agents on the Maybe* platform.