Data protection and privacy review
Map the data each integration processes and the privacy evidence needed before a rollout.
On this page
Build a data map for your integration
Kastra evaluates the context needed to apply an action policy. The data involved depends on the integration and enabled features. Start with synthetic content and review the collection scope before using production repositories, prompts, or tool arguments.
| Data path | What is processed | Review decision |
|---|---|---|
| Edge evaluation | Supported tool input, target paths, action attributes, and available prompt context sent to Kastra | Which repositories and action fields may enter the service? |
| Proxy | Supported provider request and response content; upstream credential used for forwarding | Which models, data classes, provider accounts, and output paths are approved? |
| Local MCP | Tool names, catalog context, and call arguments used for evaluation | Which servers and sensitive arguments may be connected? |
| Recon | Supported local history, redacted on the device before candidate summaries are uploaded for analysis | Which devices and history sources may participate in scanning? |
| Policy assistance | Rule descriptions and supported examples/context used by configured model services | What may be included in a draft or suggested correction? |
| Usage reporting | Supported client-reported usage events, separate from proxy metering | Which telemetry is enabled and who reviews its collection? |
Understand masking, storage, and forwarding
Stored-payload masking, encrypted-original storage, and forwarded-content redaction are separate controls. A masked decision can coexist with an encrypted original under the configured storage policy. A redacted outbound request changes what reaches the provider on that path.
Test these controls with synthetic sensitive values. Inspect the recorded view, exported material, and downstream payload where you control the receiver. Detector scope and payload limits determine what is inspected. Technical data-handling guide.
Prepare the GDPR review
Use the Kastra Trust Center as the starting point for GDPR documentation and privacy due diligence. Identify the parties’ roles for your use case, processing purposes, data categories, and instructions. Review the applicable data processing agreement and technical and organizational measures with your privacy team.
The European Commission’s data-protection obligations guidance covers transparency, safeguards and ongoing review. Have your privacy team assess the processing relationship and applicable terms for your deployment.
Confirm the deployment-specific terms
Record the agreed processing locations, relevant subprocessors and analysis providers, access controls, retention and deletion schedules, support access, and handling of data-subject or incident requests. For international transfers, have your privacy team review the applicable mechanism and contract for the actual parties and locations.
A policy’s jurisdiction attribute supplies evaluation context; it does not select a hosting region. Confirm residency in the deployment and contractual documentation. If a document is unavailable in the portal, request it from Kastra and record the open question before approving the affected data use.
Keep the review current
Revisit the data map when enabling another integration, Recon scanning, policy assistance, telemetry, or a different provider. Name an owner for retention and deletion verification. The assurance guide explains how vendor evidence and your implementation review fit together.