LLM Guardrails Alert

To generate LLM guardrails alerts, you need to create an LLM policy as demonstrated in the LLM Guardrails Policies. After creating the policy, provide a prompt in the chatbot application. The alert will then be visible in the alerts section.

Attached a dummy chatbot message demonstrating how vulnerabilities flagged by policies will be displayed.

The "View Details" feature allows users to see a comprehensive summary, including user information, detector, data, Prompt and history.

"By clicking on the three vertical dots, users can take specific actions on detected content. Options include notify, confirming a true positive detection, ignoring a false positive, resolve, or assigning to someone else. Additionally, all actions and decisions are logged, allowing users to view the complete history of the content review process."

Confirm(True Positive)

For complete instructions on managing this reason, please visit here for the full guide.

Select a option others if not reason in the list.

Ignore

Resolve