Skip to main content

Policy Actions

Block​

Blocks query from being sent to LLM/user

Sanitize​

  • Can only be used when the method is PII and applicableTo is INPUT
  • Redacts PII from prompts before sending to the LLM. Reverse maps the redacted entites in the LLM response back to plaintext before any downstream processing. In this way, the LLM never sees sensitive information.

Redact​

  • Can only be used when the method is PII
  • Redacts PII from prompts before sending to the LLM. In this way, the LLM never sees sensitive information.

Warn​

Flags internally

None​

No action taken