January 2026

Zscaler AI Guard Policy Generator

The Policy Generator has been upgraded to now support the creation of policies for Zscaler AI Guard. This improvement allows users to create and manage policies specifically designed to enforce security, compliance, or operational guidelines on Zscaler AI Guard with greater ease and efficiency.

Figure 1: Generate a New Zscaler AI Guard Policy

Report: Heatmap and Compliance Policies

Reports have been enhanced to include separate Heatmaps based on Attack Strategies, offering more granular and actionable insights.

Additionally, a new "Policies" section has been added to Target Overview Report. This section includes all compliance policy mappings that are marked for inclusion in the report (via compliance page).

Figure 2: Fail Rate Heatmaps per Attack Strategy

New and Updated Model Benchmarks

The Model Benchmarks have been updated to include results for the following newly added or revised models:

• deepseek-ai/DeepSeek-R1-0528

• deepseek-ai/DeepSeek-R1-0528-Qwen3-8B

• deepseek-ai/DeepSeek-R1-Distill-Qwen-14B

• deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B

• deepseek-ai/DeepSeek-R1-Distill-Qwen-32B

• deepseek-ai/DeepSeek-R1-Distill-Qwen-7B

• Gemini 2.5 Pro

• mistralai/Mistral-7B-Instruct-v0.3

Improvements & Tweaks

• Platform Notifications: Messages now support links.

• Report: Added links to Probe Runs, Test Cases, etc.

• Test Case: Added error test cases to CSV and JSON exports.

• Benchmark Version: Displayed on UI/API/Report.

• Probe Settings: Added a Coverage column (similar to Risk Priority).

• Benchmark Report API: Added an optional filter to exclude probes.

• AI Assets - MCP Server: Minor improvements.

• Test Case Results: Added a copy button for text messages only, while retaining formatting.