Microsoft — Volume 344, Issue 1

Microsoft

Volume 344, Issue 1 · 12 articles

In This Issue


Stop Treating Agents as One-Off Projects and Start Running Them as a Continuous Discipline (Microsoft, Thursday, November 5th, 2026)

Microsoft urges running AI agents as a continuous business discipline, citing an IDC study on AI funding, costs and governance.

An IDC InfoBrief sponsored by Microsoft found nearly 90% of organizations apply financial thresholds before approving AI initiatives, but only 17% model how costs scale in production, while about 78% prioritize cost efficiency in model selection.

86% expect financial results within six to 24 months.

Microsoft argues AI should be managed as a continuous discipline and proposes four questions for each workload: fund it, run it, improve it and prove it.

Recommendations include projecting usage at 3x and 10x, tracing spend to apps and agents, measuring cost per outcome rather than per token, and trimming context, with Microsoft Foundry positioned as the platform for this.

Read more

What's New in Microsoft Foundry: October 2026 (Microsoft, Thursday, November 5th, 2026)

Microsoft Foundry's October 2026 update adds the Copilot harness preview, GA Browser Automation, hosted-agent resilience and Decisions API.

Microsoft Foundry's October 2026 roundup introduces the Microsoft Copilot harness in public preview in Foundry Agent Service, giving prompt agents a managed loop and sandbox to read files, run code, use tools and apply skills.

Browser Automation and long-running resilience for hosted agents are generally available, and the Decisions API has come to Foundry.

The post also covers Foundry Models, speech and audio, Foundry Local and SDK changes to review before upgrading.

It recommends Foundry sessions for Microsoft Ignite, November 17-20, 2026, including a Foundry Discord Live session on November 19.

Read more

What's New in Microsoft Security: November 2026 (Microsoft, Thursday, November 5th, 2026)

Microsoft Security's November 2026 updates add AI agent permission controls, Execution Containers GA, Intune EPM and Defender changes.

Microsoft Security's November 2026 roundup covers Intune, Defender and Secure AI updates.

Dynamic Permissions scopes an agent's authority to its current task, Dynamic Agent Risk assesses whether that authority matches its behavior, and Dynamic Conditional Access can narrow permissions or require human approval.

Microsoft Execution Containers is generally available, providing OS-enforced least-privilege containment for AI agents across files, networks and processes.

Intune Endpoint Privilege Management now lets standard users change network adapter IP and DNS settings, and a Secure Now update in Exposure Management links recent threats to focused remediation initiatives.

Read more

Email Threat Landscape: Q3 2026 Trends and Insights (Microsoft, Thursday, November 5th, 2026)

Microsoft detected about 7.3 billion phishing emails in Q3 2026, with September up 41%, as QR phishing fell and CAPTCHA-gated lures rose.

Microsoft Threat Intelligence detected about 7.3 billion email phishing threats from July to September 2026, including nearly 3 billion in September, up 41% from August.

QR code phishing stayed far below its early-year peak, while CAPTCHA-gated phishing rebounded, partly through SVG attachments.

About 129 million malicious messages spoofed business-service sender domains, and many were delivered because configured exceptions overrode protections.

Campaigns included salary-review QR lures leading to device code phishing against legal firms and payroll lures delivering Agent Tesla, while Tycoon2FA-linked messages fell 99.8% after its March disruption.

Read more

Microsoft named a Leader in the 2026 Gartner Magic Quadrant for AI Agent Management Platforms (Microsoft, Tuesday, November 3rd, 2026)

Microsoft was named a Leader in Gartner's inaugural 2026 Magic Quadrant for AI Agent Management Platforms for Microsoft Agent 365.

Microsoft says it was named a Leader in the first Gartner Magic Quadrant for AI Agent Management Platforms, recognizing Microsoft Agent 365, in which more than 50 million agents have been registered.

Its approach starts with identity: Microsoft Entra Agent ID gives each agent a unique ID tied to an owner, purpose and access controls that adjust with risk signals.

Rather than a separate platform, Agent 365 extends Microsoft Entra, Defender, Purview, Intune and the Microsoft 365 admin center.

Microsoft says Agent 365 is designed to support agents from Microsoft, customers and other ecosystems.

Read more

ARTEX Abuse Expands as Cybercriminals Target Financial Sector (Microsoft, Friday, November 6th, 2026)

Microsoft says Cambodia-based Storm-3358 began using the agentic AI framework ARTEX to attack financial organizations in October 2026.

Microsoft Threat Intelligence reports that Storm-3358, a financially motivated, Cambodia-based intrusion cluster, began targeting financial organizations with the agentic AI offensive framework ARTEX in October 2026.

The group mainly targets Southeast Asian financial firms plus some large global institutions.

It combines ARTEX with automated tools to scan internet-facing systems, exploit flaws such as anonymous file upload and SSRF, and install open-source RATs including WISP, vshell and AdaptixC2.

ARTEX agents can also brute force passwords, spray credentials, abuse APIs, dump databases and exfiltrate data, and Microsoft is publishing indicators of compromise.

Read more

Threat Actors Use Agentic AI Tooling to Find and Abuse Exposed Secrets (Microsoft, Thursday, November 5th, 2026)

Microsoft observed attackers using the agentic framework ARTEX to find exposed secrets on GitHub and abuse compromised OAuth apps.

Microsoft Threat Intelligence observed threat actors using the ARTEX agentic AI framework to search public GitHub repositories for exposed credentials, OAuth application secrets and AI API keys.

The agents also analyze results and conduct follow-on attacks, in one case moving from scanning GitHub to exfiltrating data from Azure Key Vault in five minutes.

Since October 28, 2026, Microsoft saw successful service principal sign-ins from 199.19.77[.]84 using multiple application IDs, including a compromised multitenant vendor application used to reach downstream tenants.

The attackers performed cloud resource discovery, credential and data exfiltration from Key Vaults and Storage Accounts, and abuse of Azure AI services.

Read more

Microsoft Foundry Extension for Azure Developer CLI Is Generally Available (Microsoft, Thursday, November 5th, 2026)

The Microsoft Foundry extension for Azure Developer CLI reached 1.0 GA for setting up, testing and deploying hosted agents.

The Microsoft Foundry extension bundle for Azure Developer CLI (azd) has reached version 1.0 and is generally available.

It adds agent setup and local testing to azd, which provisions Azure resources and deploys code, keeping agent and deployment configuration in azure.yaml alongside the code.

Hosted agents in Foundry run a developer's own agent code on managed compute.

The walkthrough requires azd 1.37.0 or later, uv and Python 3.13, and shows initializing a Python Agent Framework sample, provisioning resources and model deployments, and optionally ejecting Bicep or Terraform files.

Read more

Predicting and Detecting Failures: How Azure Strengthens Reliability for Mission-Critical Workloads (Microsoft, Wednesday, November 4th, 2026)

Azure Failure Prediction and Detection predicts hardware failures and moves workloads, improving prevention of VM interruptions by 36%.

Azure Failure Prediction and Detection is an end-to-end system, fully operational since 2024, that uses machine learning and telemetry to predict, mitigate and remediate hardware failures before they interrupt workloads.

Microsoft says it has improved prevention of hardware-related VM interruptions by 36% across the Azure fleet and reduced hardware reboots by more than 36%.

It moves eligible workloads and notifies customers when action is needed, operating across General Purpose, Specialized Compute, HPC/AI and select Storage workloads.

SAP on Azure is used as an example of its value for mission-critical services, and the underlying system won a Best Paper Award at the 2025 IEEE Cloud Summit.

Read more

Fake-Download Campaign Uses Trojanized Installers to Deliver Monetization Payloads (Microsoft, Tuesday, November 3rd, 2026)

Microsoft Threat Intelligence details a fake-download campaign using trojanized Inno Setup installers to deliver monetization payloads.

Microsoft Threat Intelligence describes a campaign targeting users searching for cracked software, free games, activators, plugins and unofficial utilities.

Trojanized Inno Setup-style installers deliver multiple monetized payloads, called offers, in a single installation session, using per-offer callbacks to track outcomes.

Microsoft confirmed three monetization branches: routing third-party internet traffic through the device, enrolling it in a distributed computing platform and installing additional utility software, often without informed consent.

The post provides mitigations, Microsoft Defender detections, hunting queries, MITRE ATT&CK mappings and indicators of compromise.

Read more

What Exploited Vulnerabilities Have in Common: A Data-Driven Analysis of Real-World Attacks (Microsoft, Monday, November 2nd, 2026)

Microsoft analyzes which CVEs get exploited, finding about 0.5 percent of disclosed vulnerabilities are linked to real-world attacks.

Microsoft says roughly 383,000 vulnerabilities were disclosed from 1999 through September 2026, but only about 2,000 CVEs have been linked to real-world exploitation.

Attackers disproportionately target flaws enabling system takeover, especially remote code execution, privilege escalation and authentication bypass.

Because patching can take weeks or months and some systems cannot be patched, Microsoft urges defenders to prioritize by exploitation likelihood, exposure and blast radius.

Its analysis classifies each CVE by impact class and by affected platform, such as Windows, Linux, mobile, network appliances, IoT and developer packages.

Read more

Bring Agentic Coding On-Device with Foundry Local (Microsoft, Thursday, November 5th, 2026)

Foundry Local runs Qwen3.8-27B on Nvidia GPUs and connects to GitHub Copilot for private, offline agentic coding.

Microsoft's Foundry Local brings agentic coding to developer hardware by running Qwen3.8-27B locally and connecting it to GitHub Copilot through Bring Your Own Key.

Local inference keeps code on the machine, makes costs predictable and works offline.

Powered by ONNX Runtime, Foundry Local optimizes prefill, decoding, memory, long-context processing and concurrent requests on Nvidia GPUs.

In tests on an RTX 5090 and DGX Spark using real GitHub issues, it delivered higher generation speed than llama.cpp at every prompt length, with the gap widening beyond 131K tokens, and it requires a GPU with at least 24 GB of VRAM and Foundry Local CLI 0.13.0.

Read more


More from Microsoft · All section posts · This week