AI Security Auto-populate organizes supported garak results for LLM applications, agents, RAG systems, model APIs, and related integrations. Candidates enter the normal voiqq AI security project workflow and remain Pending until reviewed.
How the integration works
voiqq normalizes probe results, model and detector context, prompts, responses, severity signals, and likely OWASP AISVS requirements. Ambiguous mappings remain visible for review.
What you need
- An AI Application Security project with the target and trust boundaries defined
- Authorization to test the model, service, agent, or integration
- A supported garak report or safely configured test endpoint
- A plan for removing secrets, sensitive prompts, and unnecessary personal data
A practical workflow
- Open Auto-populate in the AI Application Security project
- Choose the supported import or configured scan method
- Review probe completion, candidate evidence, and mapping warnings
- Accept legitimate candidates into the finding workflow
- Confirm reproducibility, affected flow, likelihood, impact, remediation, and retest
What enters the project
- Probe, detector, generator, model, and test context
- Minimum necessary prompt and response evidence
- Severity signal and likely AISVS requirement
- Source metadata, remediation context, and duplicate fingerprint