Harmful Response

Monitor harmful responses from AI models to user prompts, such as self-harm and illegal activities. Track the top users triggering harmful responses, the top intentions, and the number of blocks over time. To broaden your protection, enable the Harmful Response Prevention Guardrail in Policies.

To set up what this dashboard reports on, see Guardrail: Harmful Response Prevention. For the date range and export controls that all Guardrails dashboards share, see Guardrails.