Security and privacy

Configure identity, session lifetime, connector scopes, model profiles, retention, local-content persistence, execution targets, approvals and audit export for your risk model.

Provider and service credentials remain outside customer-visible prompts. Available capabilities follow the authenticated user’s permissions and organization policy. Treat browser pages, repository files and connector results as untrusted inputs and review them before use.

Use request IDs and content-free operational telemetry for support. Do not put tokens, prompts, source contents or sensitive page data in diagnostic labels. For regulated data, confirm the contracted deployment, retention and connector terms with Pervaziv before enabling the workflow.