Violation Log
The Violation Log tracks all governance violations across your AI operations, providing a centralized view of policy breaches, blocked agents, and halted sequences.
Overview
A violation is recorded whenever the governance pipeline detects a policy breach, an unauthorized action, or an anomalous condition. The Violation Log gives you full visibility into these events so you can investigate, resolve, and prevent recurrence.
Violation List
The main view displays a list of all recorded violations with:
- Timestamp -- When the violation occurred
- Severity Badge -- Color-coded severity indicator
- Violation Type -- Category of the violation
- Agent -- The agent involved
- Summary -- Brief description of what happened
Severity Levels
| Severity | Badge Color | Description |
|---|---|---|
| CRITICAL | Red | Immediate threat to safety or compliance. Requires urgent attention. |
| HIGH | Orange | Significant policy breach. Should be investigated promptly. |
| MEDIUM | Yellow | Notable policy violation. Review at next opportunity. |
| LOW | Blue | Minor violation or informational. Track for patterns. |
Filtering
Filter violations using the controls at the top of the log:
- Severity -- Show only violations of a specific severity level
- Type -- Filter by violation type (see types below)
- Agent -- Show violations associated with a specific agent
Violation Types
| Type | Description |
|---|---|
SGP_VIOLATION | Action violated one of the 21 Sentinel Governance Protocol directives |
CHAM_VIOLATION | Action failed a CHAM policy check (confidence floor, rate limit, etc.) |
AGENT_BLOCKED | An action was submitted by an agent in a blocked state |
SEQUENCE_HALTED | A multi-step action sequence was stopped mid-execution due to a policy breach |
NO_VALID_GRANT | The agent does not have a valid grant for the requested action type |
L6_MODEL_GOVERNANCE | Model-level governance detected an issue (prompt injection, unsafe output, etc.) |
Violation Details
Click any violation to view its full record:
- Sequence Number (seq) -- Unique identifier for the violation event
- Rule Violated -- The specific policy or SGP directive that was breached
- Action Context -- Full details of the action that triggered the violation, including agent, action type, target service, and payload
- Governance Reasoning -- Explanation of why the action was flagged
- Related Audit Entry -- Link to the corresponding audit trail record
Resolving Violations
Each violation can be marked as resolved once it has been investigated and addressed. Resolving a violation does not remove it from the log -- it remains as a permanent record with a resolved status and timestamp.
TIP
Establish a regular review cadence for violations. Even LOW severity violations can reveal patterns that indicate misconfigured policies or agents that need additional constraints.
Exporting Violations
Export the violation log as a PDF or CSV file for external reporting, compliance audits, or incident investigation. Exports respect any active filters, so you can export a targeted subset of violations.
WARNING
Violations are part of the immutable audit record. They cannot be deleted or modified after creation. The resolve action only adds a resolution status -- it does not alter the original violation data.