Guardrails enhance the security and control of your application by allowing you to enable them for Assistants.
You can configure these settings in the Console when defining a new Chat Assistant or by clicking on EDIT PROMPT for an existing Chat Assistant.
When you open the Assistant Editor, the Security Guardrails section appears on the right side of the screen. From this section, you can enable one or more of the following options:
- Prompt Injection: Protects the Assistant from malicious instructions hidden in user inputs or system commands that attempt to change its behavior.
- Input Moderation: Detects and manages harmful or inappropriate user inputs before the Assistant processes them.
- Assistant Output: Helps ensure that the Assistant’s responses are appropriate and reliable, reducing biased or unsafe outputs.
Since version 2025-01.