Browse Azure News (134)

Rena Liu introduces Azure AI Speech LLM 09-2026, adding 25 languages and faster multilingual transcription. The announcement covers recognition improvements, locale selection, phrase lists, and REST API configuration, with automatic service-side deployment requiring no customer action.

Sverre Brandsberg Dahl describes how SLB agile reservoir modeling connects with Azure Data Manager for Energy to run reservoir ensembles at cloud scale. The architecture combines parallel simulation, shared Azure Files storage, OSDU-aligned data services, and Microsoft Entra ID access controls to keep modeling workflows connected to governed corporate data.

Cynthia Zanoni and Pooja Yarabothu explain how Microsoft’s PostgreSQL extension brings database exploration, SQL editing, and troubleshooting into VS Code. They show how GitHub Copilot can help interpret schemas and query plans, while emphasizing result verification and careful review of performance recommendations before making changes.

Josef Sin explains what Azure DevOps users should check following the PHP2Shell disclosures. Azure DevOps Services and Server do not use PHP to handle customer requests, but pipeline workloads, agent hosts, containers, and deployed applications may require runtime patches. The guidance distinguishes Microsoft-managed updates from fixes customers must apply themselves.

Microsoft Security Research, Mahesh Mandava, Srinivasan Govindarajan, Sagar Patil and Suriyaraj Natarajan break down the LiteLLM 1.107.0 supply-chain compromise, showing how Python startup hooks enabled credential theft without importing LiteLLM, what data was targeted, and how to investigate and recover using Microsoft Defender detections and hunting queries.

Eleanor White explains how GitHub is moving its Git storage to Azure, focusing on the last ultra-busy repositories blocked by cross-region latency. The post breaks down per-reference write quorums and a “virtual replica” design that makes Azure Blob Storage part of the durability path so reads and writes can scale independently.

jovanpop-msft introduces preview support for calling Fabric User Data Functions from T-SQL in Fabric Data Warehouse. The integration extends warehouse queries with custom Python logic for data enrichment, external API calls, and secret retrieval, with an example that calculates and stores IPv6 network prefixes for downstream analytics.

Alistair Speirs outlines how to design “sovereign AI” architectures that keep sensitive workloads under the right operational, regulatory, and security controls, while still enabling AI capabilities from edge to cloud. The post connects Azure Local, confidential computing, and Microsoft Foundry/Foundry Local into a single, verifiable foundation.

Bruno Capuano shows how to use Bring Your Own Model (BYOK) with the GitHub Copilot SDK in a small .NET 10 console app, routing inference to Microsoft Foundry or Azure OpenAI while keeping the Copilot runtime. The walkthrough covers ProviderConfig, endpoint normalization, and Entra ID vs API-key authentication.

Nandini Muralidharan announces the general availability of the Browser Automation Tool in Microsoft Foundry, explaining how agents can use managed cloud browsers (via Playwright Workspaces) to automate multi-step web workflows across public sites and private apps behind an Azure virtual network, with governance and human-in-the-loop controls.

sbaynes interviews Mercedes-AMG PETRONAS F1 lead race strategist Joey McMillan on how the team uses Microsoft Azure—especially Azure Batch—to run large-scale race simulations quickly enough to influence live, race-day strategy decisions under changing weather and track conditions.

Michael Scovetta summarizes Microsoft and NVIDIA’s guidance on managing risk when deploying open-weight AI models, focusing on how risk depends on the model, use case, deployment architecture, and safeguards, and how to connect risk categories to evaluations and controls.

Nick Brady’s September 2026 roundup covers Microsoft Foundry updates across models, agents, observability, and tooling—highlighting GPT-6 family GA, Routines and Tool Search GA, Insights preview, Model Router preview controls, and Foundry Local 2.1 bring-your-own-model support with ONNX Runtime GenAI.

Mike Hulme explains how Azure is evolving its cloud-native platform to better support AI workloads and AI agents, focusing on coordinated compute/storage/networking, production-ready open-source foundations (notably Kubernetes), cost governance with FinOps, and “agentic” operations to improve reliability and efficiency at scale.

Lim Ai Leen reports on Bar-Ilan University’s pilot of an AI agent in Moodle that helps students search, summarize, and quiz themselves on course materials using natural language, built with Azure AI Services and Microsoft Copilot Studio and run in the university’s Azure environment.

Eduard van Valkenburg explains how Microsoft Agent Framework for Python models and uses vector stores for embeddings, vector search, and agent context. The post shows how to define typed records, index and query data, expose vector search as an agent tool, and keep conversation history searchable via context providers across multiple database integrations.

Sandeep Sen announces the stable Azure MCP Server 3.0 release and explains what’s new for bringing Azure operational tooling into MCP-compatible AI agent workflows, including Terraform and Advisor tooling, more predictable client integration, and security-focused defaults for resource creation.

Jakub Oleksy reports on GitHub’s September 2026 reliability work and incidents, including database pressure reductions, overload controls, added headroom for GitHub Actions, expanded failure containment for pull requests, and progress moving production workloads and traffic to Azure.

Joe Filcik and Colin Murphy explain how Azure Content Understanding uses grounding and field-level confidence to make document extraction results inspectable and actionable, then show how to calibrate confidence thresholds (per document type and per field) to route work in human-in-the-loop review workflows.

Microsoft Threat Intelligence and Microsoft Digital Crimes Unit detail how a Chinese-language phishing-as-a-service ecosystem (Storm-2990 and Storm-3119) monetizes compromised Microsoft 365 and Azure for Education tenants, then provide practical mitigation guidance and Microsoft Defender detections to help organizations reduce phishing-driven account takeover risk.

sbaynes outlines how Microsoft is scaling Azure’s AI infrastructure end-to-end, covering custom silicon (Maia and Cobalt), networking for large AI clusters, and datacenter power and cooling redesigns. The post also highlights hardware-rooted security work like Azure Integrated HSM and the open-source Tessera Integrated HSM initiative.

sbaynes breaks down the Azure Maia 300 system architecture and the design choices Microsoft made to optimize cloud-scale LLM inference, focusing on dataflow, memory locality, low-precision formats, and rack-to-superpod networking to improve tokens-per-watt and tokens-per-dollar.

mk_sunitha introduces Fabric App functions (preview), a TypeScript-based serverless capability for Fabric Apps that lets builders run secure server-side logic, use application authentication to access downstream resources, and manage secrets safely. The post also shows how to scaffold, debug, and deploy functions, including using GitHub Copilot in VS Code.

Mark Russinovich and Saurabh Dighe announce Project Tessera, an open-source evolution of Azure Integrated HSM, and explain how its integrated HSM model and protocols aim to improve key protection, interoperability, and post-quantum readiness—plus how the same firmware stack can extend into confidential computing environments.

Chu Lahlou explains how Azure Content Understanding prebuilt analyzers go beyond a schema by packaging domain knowledge with grounding and confidence signals, so document extraction results are safer to automate. The post also shares benchmark results for Advanced Contextualization in Content Understanding 2.0 Preview, covering quality, modeled cost, and latency.

sbaynes announces Microsoft’s plan to scale the Azure Maia accelerator system to multi-gigawatt deployments starting in 2027, outlining how Maia 200 is already running production inference and what architectural changes in Maia 300 target long-context, latency-sensitive, and agentic workloads.

Tamir Dresher explains where Durable Task (and Durable Task Scheduler) actually helps in production: coordinating long-running workflows that need durable waits, replay, and recovery. Using five concrete scenarios—including Azure tenant provisioning and an AI-assisted incident workflow—he shows what durability solves and what your app still must own.

Tsuyoshi Ushio introduces Dynamic Workflows for Azure Functions Hosted Skills, a preview feature that lets an LLM generate a typed DAG plan once and then run it durably via Durable Functions. The post explains the runtime model, safety checks, and when this approach reduces token usage and latency for event-driven apps.

Chu Lahlou compares Azure Document Intelligence (ADI) and Azure Content Understanding (ACU) for real-world document processing, focusing on when to use purpose-trained document models versus generative, schema-driven extraction. The guide covers OCR/layout overlap, custom extraction trade-offs, RAG-ready output, multimodal analysis, and a practical evaluation checklist.

Rani Borkar explains how Microsoft applies agentic AI across Azure’s hardware and infrastructure lifecycle—from demand planning and supply chain through datacenter fleet operations—to shorten decision cycles, improve reliability, and build a closed-loop learning system that keeps human judgment in control.

Brian Celenza explains how GitHub is redesigning its Git storage architecture to handle “agentic” development at extreme scale, where pushes, merges, and CI fan-out happen concurrently and continuously. The post outlines the bottlenecks in today’s replica-based design and the principles behind a new storage/compute split.

Alex Weininger introduces the Canvas authoring plugin, which guides GitHub Copilot through the native create-canvas workflow to generate an Azure canvas app based on Microsoft Canvas Toolkit, including a read-only starter and patterns for safe, explicitly scoped Azure operations.

Freddy Dezeure and Sesha Mani explain how frontier AI models are changing vulnerability management, and what CISOs can do to keep patching both fast and correct. The post highlights Microsoft’s use of AI “harnesses” for scanning and triage, and points to Baseline Security Mode and secure-by-default controls to reduce blast radius when patches can’t land in time.

Victoria Hall introduces Azure Functions Agent bindings (preview) for Python, showing how to add bounded agentic reasoning via Microsoft Agent Framework while keeping deterministic function code in control. The post walks through project structure, package setup, local settings, and examples for both direct HTTP calls and Durable Functions orchestrations.

James Rempt explains how Azure credits included with eligible Visual Studio Subscriptions can be used as a personal Azure sandbox for learning and experimentation, including how the monthly credit works, what happens when you hit the spending limit, and key constraints around non-production use and service eligibility.

Lily Ma introduces Azure Functions managed connectors (public preview) and shows how connector triggers and typed clients let functions react to events and call actions across services like SharePoint and Teams without hand-rolling webhook registration or OAuth token handling.

Allison rounds up the latest GitHub Copilot updates for the week, including new model options, HydraFusion model coordination in preview, reusable dynamic workflows, and “computer use” for automating desktop apps. The post also highlights new Azure canvases in the Copilot app and experimental VS Code agent improvements.

Kristen Womack recaps the September 2026 Azure Developer CLI (azd) releases, covering extension dependency handling, versioned gRPC extension contracts, new azure.yaml project layering, per-phase concurrency controls for provisioning and deployment, and authentication host connectivity via local socket transports.

Chu Lahlou and Cha Zhang explain why enterprise GenAI systems still need a dedicated content-extraction layer, even with stronger LLMs. They break down the operational gaps of “LLM-only” extraction and position Azure Document Intelligence and Azure Content Understanding as complementary tools for grounded, auditable, multimodal workflows.

Justin Bettencourt rounds up the September 2026 Azure SDK releases across .NET, Java, JavaScript/TypeScript, Python, Go, Rust, and C++, calling out notable updates like Azure AI Search preview API support, a stable Rust SAS crate, and new Entra ID authentication libraries for PostgreSQL.