AI Security
Search K

Appearance

June 17, 2026 - SaaS - v10.42.6 ​

This release includes updates in the following areas for F5 AI Security:

AI Guardrails ​

PII and language integration ​

Spanish (es-ES) is now avaialble by default. In the UI, see AI Guardrails > Guardrails and look for available PII packages. You'll also need to install the default English (US) PII package.

If you have an on-premises installation, you may have to reinstall AI Security by redeploying to the cloud, or using Helm.

Streaming support ​

The F5 AI Security platform sits between your application and upstream providers such as OpenAI, Anthropic, and Google.

Fingerprints for analysis ​

You can now resize the fingerprints panel in the UI. It'll help you navigate and read agentic traces. For more information, see Generate fingerprints for analysis.

AI Red Team ​

New attack pack ​

AI Red Team now includes a new attack pack: Cyberpunk stories.

Cyberpunk stories tricks AI models into producing harmful content. It disguises requests as a literary analysis task set within a cyberpunk story. It works by asking the model to break down the narrative structure of a story that actually contains dangerous instructions.

Refusal checker ​

You can now use the API add phrases to the refusal checker to make it more accurate to the specific model, application, or use case. We've created a new refusal-catalogs/ endpoint.

Platform ​

Improved UX when you invite a new user. The panel is bigger. If the default role is a custom role, it appears at the top of the custom list.

For on-premises installations, we now also support Oracle Kubernetes Engine (OKE).

Supported deployments now include Amazon EKS, Azure AKS, Google GKE, Red Hat OCP, and Oracle OKE.

Resolved issues ​

This release fixes the following issues:

  • API Reference Docs: Now includes sample responses. cURL command samples now properly preface each token with Bearer
  • Text search: You can now sort search results and audit logs by timestamp
  • UI. When update is available, info text no longer disappears
  • Policy management guardrail now displays details
  • Multi-turn attacks now returns correct prompt
  • Call to POST /admin/import with clear_existing=True now works
  • Custom roles now supports personal API tokens
  • Filtered raw data now returns results
  • When switching panel tabs, the Attack pack name is retained
  • Roles: Option to set default no longer appears before new role is saved
  • Roles. Users can see all of their roles when navigating to Settings > Account
  • Campaigns. Validation of attack pack file is required

Doc updates ​

We have published the following updates to our product documentation since the last release on May 21, 2026:

Updated at:

© 2026 F5, Inc. All Rights Reserved