AI Violations
AI Violations lists every place where AI Guardrails found something: which agent, which category, which values, and in which trace. Rows come from both guardrail modes:
- Scheduled Detection results, shown with source Platform.
- Real-time Enforcement checks that found a violation, shown with source SDK.
Enterprise feature. Open it from Governance → AI Violations. The list contains flagged prompt and response content, so it requires the PII data permission (pii:read) in Roles & Access. The detected values themselves are hidden from users who are not allowed to see trace input and output.
What each row shows
| Column | Description |
|---|---|
| Timestamp | When the trace or check happened |
| Name | The agent. For Platform rows it links to the trace. |
| PII, PCI, PHI, Ethics/Bias, Safety | Which categories were found |
| Detections | The labels that were found, for example email, credit card number or AWS key, and the names of matching custom detectors |
| Detected At | When the detection ran |
| Source | Platform (Scheduled Detection) or SDK (Real-time Enforcement) |
| Result | For SDK rows: Pass (Audit Only), Warning, Blocked or Sanitized |
Filtering
- Detection type: All Detections, PII, PCI, PHI, Ethics & Bias, or Safety.
- Agent: all agents or one agent.
- Time range: Last 24h, Last 7 days (default), Last 14 days, Last 30 days, or All time.
You can also open the list already filtered from other views, for example from an agent's governance profile.
Violation details
Select a row to open its details:
- Each detected value with its label and confidence score, highlighted where it appears in the input or the output.
- For Ethics & Bias and Safety, the category that was flagged, for example S10 Hate Speech, Gender Bias, or Jailbreak / Prompt Injection.
- For SDK rows, the result, the direction (input or output) and the risk level.
- For Platform rows, View full trace details opens the complete trace.
- For each detected label, the analyst marks recorded so far.
Mark true and false positives
Analysts record whether each detection was right, so noisy detectors can be tuned from evidence.
Open the detection
Select a row to open its details.
Mark the label
Next to a detected label, select Mark True Positive or Mark False Positive.
Add a note
Write a short note explaining the decision, for example "internal test card number", and save.
Each mark records the detection, the label, the decision, the analyst and the note. Marks feed the Detector Accuracy report.
Detector Accuracy report
Select Detector Accuracy on the AI Violations page to see, for each detector label, built-in or custom:
| Column | Description |
|---|---|
| Detector | The label or custom detector name, with its category |
| Detections | How many detections it produced |
| True Positives | Detections analysts marked as true positives |
| False Positives | Detections analysts marked as false positives |
| False-Positive Rate | False positives as a share of the detections analysts have marked |
The report uses the page's Agent and Time range filters. Use it to find detectors that need a higher confidence threshold, a narrower label selection or a tighter custom pattern in AI Guardrails.
Scheduled detection
Each agent's policy in AI Guardrails sets its own schedule: Hourly, Daily or Weekly. Each run scans the agent's traces received since the previous run, for the categories and labels selected in that policy. Change the schedule, labels or confidence threshold in AI Guardrails.
Incidents
Scheduled detections open incidents automatically when these integrations are set up:
- ServiceNow: when the ServiceNow integration is connected with incident creation turned on, PII, PCI, PHI, ethics and bias, and jailbreak detections open ServiceNow incidents.
- PagerDuty: when the PagerDuty integration is connected with incident creation turned on, and PagerDuty Alerting is on in the agent's policy, detections open incidents on the selected PagerDuty service with the selected severity.
Real-time Enforcement records appear in this list but do not open incidents.
Violations can contain the actual sensitive values that were detected. Limit the pii:read permission to the people who need it.
Next steps
- AI Guardrails: choose which categories run, and how.
- Drift Alerts: detection rates are compared with each agent's declared controls.
- Risk Dashboard: see risk across all agents.