2026-10
2026-10
This release tracker is LLM-curated and based on the official Microsoft product sources listed below. It provides an architect-grade summary of recent features, changes, and announcements. Always verify critical details against the official documentation. List of all raw markdown files for the releases are at https://github.com/pisinger/pisinger.github.io/tree/main/_ms_release_radar
Microsoft Security Release Radar - October 2026
Windowed security release notes for October 2026 across the Microsoft security and adjacent platform stack. Items are classified with the authoritative section status markers, and each row carries a lifecycle type (GA, Preview, Security, Deprecation, New/Updated). Security (🔴) is used only where the source marks an item as security/CVE work.
🔌 Azure API Management
| Indicator | Feature | Type | Description |
|---|---|---|---|
| 🟡 | Public Preview: Microsoft Agent 365 integration with Azure API Management | Preview | Public preview (availability 2026-10); connects centralized governance in Agent 365 with runtime enforcement for MCP servers and tools, including tool-level blocking in the AI Gateway tier and Defender/Purview evaluation of tool requests. Reference: https://azure.microsoft.com/updates?id=574204 |
☁️ Microsoft Defender Cloud Apps
| Indicator | Feature | Type | Description |
|---|---|---|---|
| 🔵 | Automatic enablement of SaaS accounts integration with Identity inventory | New/Updated | Starting October 15, 2026, the SaaS accounts integration between Defender for Cloud Apps and Identity inventory transitions from opt-in to automatic enablement for eligible tenants. |
🛡️ Microsoft Defender Cloud
| Indicator | Feature | Type | Description |
|---|---|---|---|
| 🟡 | New Defender server security experience | Preview | Public preview (October 6, 2026); brings together cloud and endpoint data from Defender for Endpoint and Defender for Servers in the Defender portal, with a combined server asset page and Defender for Cloud offboarding control. |
| 🟢 | KSPM misconfiguration recommendations moving to controller-level scope | GA/FIX | GA (October 5, 2026); KSPM recommendations now align to Kubernetes workload controllers (such as Deployment or StatefulSet) rather than individual running container instances, changing affected recommendations and Secure Score impact. |
| 🟢 | On-demand malware scanning supports scanning specific blobs, files, containers, and file shares | GA | GA (October 5, 2026); targeted scanning of specific storage items via REST API filters instead of scanning an entire storage account. |
🆔 Microsoft Defender Identity
| Indicator | Feature | Type | Description |
|---|---|---|---|
| 🔴 | New Defender for Identity security alerts | Security | October 2026 adds alerts for Entra ID, Active Directory, and other identity providers, including Kerberos diamond-ticket privilege escalation, a CVE-2025-33073 exploitation attempt, AD CS abuse (ESC1/ESC8/ESC11), CA certificate/key theft, and multiple Okta privileged-account detections. |
| 🟢 | Sensor v3.x activation without Defender for Endpoint | GA | GA for eligible domain controllers on Windows Server 2019 or later; supports only new v3.x deployments, with no v2.x migration path. |
📱 Microsoft Intune
| Indicator | Feature | Type | Description |
|---|---|---|---|
| 🔵 | Remote Help for Windows version available (5.2.1210.0) | New/Updated | [Windows] Remote Help for Windows version 5.2.1210.0 adds stability and observability improvements. |
| 🔵 | Intune security baseline for Windows 11, version 26H2 | New/Updated | [Windows] New baseline reflecting current Microsoft security recommendations with new settings and updated defaults; existing baseline profiles do not update automatically. |
📧 Microsoft Defender Office 365
| Indicator | Feature | Type | Description |
|---|---|---|---|
| 🔵 | Expanding user reporting in Teams to include meetings | New/Updated | Users can report scheduled or Meet now Teams meetings, or individual meeting participants, as a security concern during or after the meeting, with metadata routed per the Teams user reporting settings. |
🧠 Microsoft Copilot Studio
| Indicator | Feature | Type | Description |
|---|---|---|---|
| 🟢 | Foundry IQ integration with Copilot Studio | GA | Reusable Foundry IQ knowledge bases can ground multiple agents, with retrieval scope per agent, citations, and secure private-networking connectivity. |
| 🟢 | Review panel for agent readiness | GA | Surfaces blockers and warnings throughout agent authoring, flags organizational policy restrictions, and warns when an agent is unevaluated or affected by a model change. |
| 🟡 | Apps in Copilot Studio | Preview | Makers can describe an app in natural language and generate a draft app alongside agents and workflows, retaining access to the underlying code when needed. |
| 🟡 | Microsoft Copilot Managed Runtime | Preview | Microsoft-operated hosting, identity, governed data access, lifecycle management, and Git-backed versioning for apps created across Copilot Studio, Copilot Cowork, and Copilot Code. |
| 🟡 | Hooks for agents (GitHub Copilot harness) | Preview | Adds deterministic control at lifecycle events such as session start, tool calls, and errors, while preserving agentic reasoning flexibility. |
| 🔵 | Expanded agent evaluations and Custom Graders Library | New/Updated | Evaluation now covers automation and individual AI nodes, adding Task Completion, Tool Accuracy, Safety, and Latency measures plus reusable tenant-specific custom graders and a new Evaluation Viewer role. |
| 🔵 | Centralized plugin registry | New/Updated | A shared catalog for discovering and governing plugins, skills, connectors, and agents, with Copilot Studio support rolling out over the coming weeks. |
🐙 GitHub Security
| Indicator | Feature | Type | Description |
|---|---|---|---|
| 🔵 | Purpose-built model for leaked secret detection | New/Updated | A fine-tuned model for AI secret detection across secret scanning alerts, push protection, and Copilot security reviews; existing AI-detected Password alerts upgrade automatically. |
| 🟢 | Local sandboxing for GitHub Copilot now generally available | GA | GA in Copilot CLI, the Copilot app, and VS Code Agent Host sessions; powered by Microsoft eXecution Container (MXC) with enterprise-managed policies that developers cannot weaken. |
| 🔵 | Audit GitHub Code Quality finding dismissals and reopenings | New/Updated | New audit log events record manual dismissals and reopenings of Code Quality findings, including actor, repository, rule, and affected finding identifiers. |
| 🔵 | Secret scanning blocks Anthropic keys by default, adds Salesforce | New/Updated | Adds Salesforce as a partner and detects new Anthropic and Cloudsmith secret types, with Anthropic keys blocked by push protection by default. |
| 🔵 | Improved detection for withdrawn Dependabot alerts | New/Updated | A withdrawn action is added to the dependabot_alert webhook, and REST requests for withdrawn alerts now return 410 Gone instead of 404. |
| 🟡 | Code scanning AI Scan metrics in security overview | Preview | Org and enterprise admins can view AI Scan findings, remediation rate, trends, and autofix usage via new CodeQL and AI Scan tabs. |
| 🔵 | Code scanning AI Scan enablement status in security overview | New/Updated | Coverage view shows AI Scan enablement per repository, with new filters and a CSV export column for tracking adoption. |
| 🔴 | Rate limits for private vulnerability reports | Security | Daily per-user rate limits on new private vulnerability reports, with repository-level overall limits and a trusted-reporter allow list. |
| 🔴 | Structured forms for private vulnerability reports | Security | Reporters must complete structured fields (summary, details, proof of concept, impact), with customizable forms, optional CWE requirement, and AI-use disclosure. |
| 🔴 | Scheduled code scanning skips inactive repositories | Security/FIX | Weekly scheduled scans now begin only after a push or pull request triggers analysis, preventing dormant repositories from being scanned due to setup actions or language changes. |
| 🟡 | Push attribution for commits in limited public preview | Preview | GitHub durably records the authenticated pusher for each commit and adds a Committer must match pusher rule to close commit-identity spoofing. |
| 🔴 | Anonymous organization-level registries for Dependabot | Security | Adds an Anonymous authentication method for organization-level private registries reachable without credentials. |
| 🔵 | Secret scanning adds detectors for Lovable, Supabase, and more | New/Updated | Adds partner/detector coverage for Lovable Labs, Pydantic Services, and Supabase secret types. |
| 🔵 | CodeQL 2.27.2 improves C++, Go, Rust, and JavaScript analysis | New/Updated | Adds a C++ ECMAScript std::regex parser, Go/Rust/JavaScript modeling improvements, macOS 27 build limitations, and query/CLI fixes. |
| ⚫ | Unvalidated npm trusted publishing configurations now expire | Deprecation/FIX | Configurations expire 48 hours after creation and can no longer authorize publishing; issue_comment tokens are also rejected. |
| 🔵 | npm staged publishing now supports creating new packages | New/Updated | npm stage publish can now create new scoped/unscoped and private packages, entering the staged queue for maintainer promotion. |
| 🔴 | New fields for SecurityAdvisory GraphQL API | Security | Adds cveId, sourceCodeLocation, githubReviewedAt, nvdPublishedAt, and repositoryAdvisoryUrl fields plus severities and isWithdrawn query filters. |
| 🔴 | Confidential comments on repository security advisories | Security/FIX | Maintainers can post confidential comments visible only to users with write access; views are recorded in the audit log. |
| 🟡 | Repository security advisory comments API in public preview | Preview | REST endpoints to list, get, add, and edit advisory comments, with comment counts on advisory responses. |
Top 5 Action Items
| Priority | Action | Due | Affected Product(s) |
|---|---|---|---|
| 🔴 High | Apply the Windows 11 26H2 security baseline to target device profiles | This month | Microsoft Intune |
| 🔴 High | Recreate unvalidated npm trusted publishing configurations before the 48-hour expiry | Immediate | GitHub Security |
| 🔴 High | Tune SOC detection content for the new Defender for Identity alerts | This month | Microsoft Defender Identity |
| 🟡 Medium | Validate Secure Score and recommendation changes from KSPM controller-level scoping | This month | Microsoft Defender Cloud |
| 🟡 Medium | Evaluate local sandboxing policies for GitHub Copilot agent workflows | This month | GitHub Security |
Security Architect Observations
- KSPM recommendations moving to controller-level scope will shift recommendation counts and Secure Score; refresh posture baselines, dashboards, and governance reporting to avoid false trend signals.
- Automatic enablement of the Defender for Cloud Apps SaaS accounts integration with Identity inventory changes identity governance and data-flow posture; confirm role-based exclusions (user group admin / app-instance admin) and review coverage assumptions.
- Push attribution plus the “Committer must match pusher” rule materially strengthens supply-chain identity assurance; plan ruleset rollout for high-value repositories and update branch-protection policy documentation.
- Local sandboxing for GitHub Copilot (MXC) introduces a portable execution boundary for agentic tooling; align enterprise-managed sandbox policies to your least-privilege and network-egress standards.
- The new Defender server security experience converges cloud and endpoint data and enables automatic server offboarding when the Defender for Servers plan is turned off; validate RBAC, device groups, and cloud scopes before adoption.
- Copilot Studio managed runtime, hooks, and the plugin registry add new governable surfaces for AI-built apps and agents; enforce evaluation, review-panel, and registry controls as part of AI platform governance.
Security Operations Observations
- The new Defender for Identity alerts (Entra ID, AD, Okta, AD CS/Kerberos) require SOC triage tuning and likely new analytics; watch for false positives in AD CS and Kerberos scenarios and map them to existing incident workflows.
- Confidential advisory comments and structured vulnerability report forms change vulnerability intake and coordination; update SOAR playbooks and integrations that read the repository advisory REST/GraphQL APIs.
- Withdrawn Dependabot alerts now emit a webhook “withdrawn” action and return 410 Gone; update automation so withdrawals are handled rather than treated as errors.
- Scheduled code scanning and Code Quality now skip inactive repositories, so fewer unexpected weekly scans and findings should be expected; adjust coverage/health monitoring accordingly.
- New secret scanning detectors (Anthropic, Salesforce, Cloudsmith, Lovable, Supabase, Pydantic) will raise alert volume and push-protection blocks; tune allow lists and developer guidance.
- AI Scan metrics and enablement status in security overview improve adoption tracking but are observability-only; pair them with alerting and remediation SLAs for AI-generated findings.
References
This post is licensed under CC BY 4.0 by the author.