Release 2.4.0 · microsoft/zerotrustassessment

Release roll-up introducing a comprehensive AI security pillar and an Infrastructure pillar, plus correctness and performance fixes.

Install from PowerShell Gallery:

Install-Module -Name ZeroTrustAssessment -RequiredVersion 2.4.0

🤖 AI pillar assessments

AI Threat Detection

61002 — Microsoft Sentinel onboarded on at least one Log Analytics workspace
61016 — Entra ID Protection risk events flowing to the Sentinel workspace
61018 — Purview Information Protection data connector enabled on Sentinel
61021 — Microsoft 365 Copilot data connector enabled on Sentinel
61022 — Defender for AI Services enabled on every subscription hosting Azure OpenAI / AI Services
61024 — Defender XDR (unified) data connector enabled on Sentinel

AI Cloud Posture

AI Authentication & Access

61006 — AI administrative roles have assigned principals
61009 — Conditional Access covers both agent identities and agent users
61011 — Require users to use Entra ID auth to interact with agents
61012 — Risk-based Conditional Access blocks risky agent identities
61013 — Identity governance for agents (sponsors, entitlement management, lifecycle automation)
61008 — Agent identity lifecycle tagging (custom security attributes present)

AI Inventory & Lifecycle

61005 — Copilot agents are discoverable in the Agent Registry
61014 — Agent identities & blueprint principals have technical owners; no disabled agents remain

Global Secure Access

25415 — AI Gateway protects enterprise generative AI apps from prompt-injection attacks

Data security & information-protection checks in the AI pillar

35003–35008 — Sensitivity label configuration, publishing, SharePoint/OneDrive labeling, PDF labeling, IRM, default library labels
35010–35017 — Double Key Encryption, super-user membership, container labels, encryption labels, attachment inheritance, mandatory & default labeling
35019–35025 — Auto-labeling (config, enforcement, SharePoint/OneDrive), on-demand scans, OCR, Azure RMS activation & licensing
35030–35036 — DLP policies, Adaptive Protection, custom SITs, Exact Data Match, named entities, trainable classifiers
35037–35041 — M365 audit logging, Insider Risk for risky AI usage, communication compliance for Copilot & enterprise AI tools, browser DLP for AI apps via Edge for Business

50001 — Surfaces Microsoft Defender for Cloud recommendations from the connected/scanned environment in the report's Infrastructure view.

Assessment results now project onto Zero Trust Workshop tasks and emit a ZeroTrustWorkshop.json for direct import into the Workshop app (best-effort; never fails the run).

Added scopes required by the new AI checks (Lifecycle Workflows, privileged assignment/eligibility schedules, Copilot packages).

#1194 — Test 21835: performance, correctness & reporting cleanups for the emergency-access account check.
#1274 — Test 24552: explicit null guard so firewall status is detected correctly.
#1264 — Test 25392: now correctly reports "skipped" when no private-access connectors are installed.
#1235 — HTML report crashed on render despite a clean run; rendering safety fixes applied.
#1193 — Use the system browser (-UseSystemBrowser) for SharePoint Online interactive auth in Connect-ZtAssessment.
#687 — Truncate large results tables (applied to 61014) to keep the report readable.
21816 — PIM privileged-role assignment check now accounts for JIT access into role-assignable groups.