Content scanning
Scan at configured policy boundaries
Evaluate connected prompts and responses for sensitive content and configured risk signals. Detection and enforcement depend on the enabled policies, integration and scanner coverage.
Implementation review
Request and response paths
Choose which inputs and outputs pass through scanning, and test them with representative content.
Policy decisions
Configure supported block, warn or log behavior at the integration point. Review what happens if the policy service is unavailable.
Detection limits
No scanner detects every sensitive value or prompt injection. Combine scanning with access controls, approved endpoints and accountable review.
Product scope
Zentinelle AI is an independently adoptable product with an MIT-licensed core. Support and implementation are scoped separately. Workbench seats do not automatically include every agent integration or operating service.
Review the events and enforcement available in the chosen integration before relying on a control. Workload location and model destinations remain separate architecture decisions.
Review current security and compliance information
