AI Security
Search K

Appearance

CAI deploy application v3.8.0 (2026-09-22) ​

To see how you can use the CAI deploy application, read Deploy AI Security to the cloud.

WARNING

This document includes references to AI Remediate, which is not generally available yet. Contact F5 sales if you are interested in joining the early access program.

Features ​

  • F5 AI Remediate: Add support for deploying the Remediate model
    • AI Remediate automatically fixes AI model outputs flagged during red team testing. The application provisions a dedicated AI Remediate GPU node group on AWS, Azure, Google Cloud Platform, and Oracle Cloud Infrastructure (OCI).
  • GPU node groups: Increase the root disk size to 200 GB for the F5 AI Red Team and AI Remediate node groups
    • Set AWS gp3 volume throughput to 500 MB/s for the scanner, AI Red Team, and AI Remediate node groups.
  • Installer guidance: Present resolved deployment choices and emphasize operations that can significantly affect a deployment
  • Python: Add Python 3.14 support, upgrade deployment dependencies, and refresh Pulumi test coverage

WARNING

For existing deployments, run --preview-only before you apply this release. Updated Pulumi providers might produce infrastructure changes or replacement plans. Review the preview carefully before you approve an update.

Resolved issues ​

This release fixes the following issues:

  • Configuration: Preserve custom repository settings
  • Installer output: Stop printing asyncio Event loop is closed tracebacks after successful deployments
    • Continue to report finalizer errors that require action.
  • Ctrl+C handling: Exit with code 130 after canceling in-progress Pulumi resource operations
    • Press Ctrl+C a second time to stop immediately. This behavior helps prevent orphaned cloud resources.
  • Docker: Install google-cloud-cli and google-cloud-cli-gke-gcloud-auth-plugin
    • Replace the retired google-cloud-sdk apt package and support GKE authentication from the container.
  • Azure: Keep the cache prompt independent of the ingress configuration
  • Deployment endpoints: Use resolved stack outputs to prevent post-deployment runtime warnings
  • Docker: Update and optimize the Dockerfile
  • Runtime security: Improve certificate validation, container execution, temporary file permissions, and diagnostic log handling
  • PyApp: Build the Linux executable with Python 3.14
  • Providers: Restore infrastructure component compatibility and stabilize provider-focused test coverage for supported upgrades

Updated at: