Skip to main content
GuardAI
Security & Privacy

How we handle your data

GuardAI sits between your application and your model provider. This page describes what happens to data in that path and where the system can run.

Data in transit

Prompts and responses

Prompts and model responses pass through GuardAI for policy evaluation. By default, content is processed in memory and not written to disk or retained after the request completes. If your organization enables audit logging, request content is stored in your designated storage — we do not retain copies.

Model provider terms

GuardAI forwards requests to your chosen model provider. Data handling between your organization and that provider is governed by your existing agreement with them. GuardAI does not change those terms.

Defense layer processing

GuardAI's defense layer uses Google Vertex AI to evaluate prompts and responses against your configured policies. This applies regardless of where GuardAI is deployed — managed cloud, your VPC, or on-premises. Prompt and response content sent through the defense layer is subject to Vertex AI's data governance policies.

Summary

  • Content processed in memory by default
  • No retention after request completes unless logging is enabled
  • Audit logs stored in your designated storage
  • Defense layer uses Vertex AI (all deployments)
  • Model provider terms for your chosen LLM unchanged
  • TLS in transit between all components

Where GuardAI runs

The managed service runs on Google Cloud Platform. For organizations with stricter data residency or isolation requirements, GuardAI can also be deployed in your own infrastructure. All deployments require outbound connectivity to Vertex AI for defense layer processing.

Managed Cloud

GuardAI-hosted on GCP. Fastest path to production. We manage infrastructure, updates, and availability.

Your Cloud

Deploy the GuardAI container in your own VPC. You control network boundaries and data residency. Requires outbound access to Vertex AI.

On-Premises

Run the GuardAI container on your own hardware. You manage the host environment. Requires outbound access to Vertex AI for defense layer processing.

Questions about security or data handling?

We can walk through deployment options, data flows, and operational fit for your specific use case.