Browse Artificial Intelligence News (375)

Allison announces general availability of assisted permissions in GitHub Copilot CLI, GitHub Copilot app, and Visual Studio Code. A language model evaluates command, file-path, and URL requests, automatically allowing those classified as clearly safe. Enterprise administrators can control available permission modes and set defaults for new sessions.

Alper explains how Microsoft Fabric’s preview React visualization libraries support analytics apps built with AI coding agents. Using Vega-Lite specifications and shared data, theming, and interaction contracts, the components provide reusable charts and grids with formatting, scrolling, tooltips, and consistent styling rather than requiring each app to recreate these behaviors.

Allison announces GPT-6.1 Terra’s general availability in GitHub Copilot for everyday agentic coding and terminal tasks. The model uses provider list pricing under usage-based billing, with gradual rollout across supported clients. Business and Enterprise administrators can control access through model policies and default enablement settings.

sbaynes describes how Microsoft combines AI agents, testing, static analysis, and formal verification to transform large codebases. The article covers Windows toolchain migration, C/C++ modernization into Rust, and specification-driven validation, explaining how small expert teams review shared knowledge and system properties rather than every generated change.

sbaynes describes using the Microsoft Discovery app to adapt insect-monitoring systems for the Asian citrus psyllid, a pest that spreads citrus greening disease. The case study covers grounded AI agents, experimental design, permitting, and synthetic biometric data for preliminary edge classifiers, with scientists reviewing decisions throughout the workflow.

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.

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.

Ben Steenhoek and Ulugbek Abdullaev explain how configurable eagerness lets paid GitHub Copilot subscribers adjust inline suggestions in VS Code. They describe the low, medium, and high modes, the training approach behind them, reward-hacking mitigations, and experimental results showing how suggestion frequency and size change.

Allison announces an update to the GitHub Copilot usage metrics API that adds a new totals_by_client breakdown, letting enterprises and organizations see request, session, and token usage coming from tools outside the IDE (like OpenCode, Copilot CLI, VS Code, and JetBrains).

Allison announces an updated GitHub app for Microsoft Teams that combines GitHub notifications with GitHub Copilot-powered actions, letting teams respond to issue and pull request updates by mentioning @GitHub to investigate, fix bugs, run tests, and open pull requests from within Teams.

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.

sbaynes introduces Microsoft research on evaluating step-by-step clinical diagnostic reasoning with SDBench and the Microsoft AI Diagnostic Orchestrator (MAI-DxO), including accuracy and cost results across NEJM and MIMIC-IV-ED cases and plans to open-source parts of the benchmark for reproducible evaluation.

Allison summarizes the September 2026 update for GitHub Copilot in Visual Studio 2026, including PR exploration via the Git agent, preview BYOM support (with a breaking change to the new Agent harness), and a new flow to fix NuGet Audit vulnerabilities directly from the Error List.

Allison announces an update to the Copilot usage metrics API that reports configured names for custom skills, custom agents, MCP servers, slash commands, and plugins (instead of grouping them as “other/custom”), making it easier for admins to track adoption and usage trends across an organization.

sbaynes introduces Flint-interactive, a language that helps AI agents generate interactive data visualizations (not just static charts) using high-level semantic events and update operators, so users can click, filter, zoom, and link interactions across charts, text, tables, and app state.

Allison announces general availability of a global default policy for GitHub Copilot features on Copilot Business and Copilot Enterprise, explaining how unconfigured settings will be enforced during the rollout and where admins can review and override defaults.

MichalBar introduces a preview feature in Microsoft Fabric Real-Time Dashboards that uses Copilot to generate purpose-built custom visuals. By describing an operational question in natural language, dashboard authors can have Copilot create both the KQL query and an HTML-based visualization directly as a new dashboard tile.

Allison announces a new “Industry comparison” view in the GitHub Copilot impact dashboard, letting opted-in enterprises benchmark agentic adoption and leverage against aggregated peer data, with controls for policy-based participation, daily refresh, and data residency boundaries.

sbaynes reports on Microsoft Research work proving the KLS conjecture, a long-standing result in high-dimensional geometry with implications for sampling, optimization, and randomized algorithms. The post also describes a human-directed, agentic AI workflow used to explore candidate proof paths and support derivations and verification.

Microsoft Security Research and Microsoft Threat Intelligence detail how Storm-0940 used the ARTEX agentic AI framework to accelerate a ConnectWise supply chain compromise, then abused OAuth application identities to access Microsoft 365 tenants via app-only authentication and Microsoft Graph, with concrete mitigation steps, detections, IOCs, and hunting queries.

prl introduces the preview of Fabric Migration Assistant for migrating Teradata workloads to Fabric Data Warehouse, focusing on translating metadata and SQL/BTEQ into Fabric-compatible T-SQL, reviewing exceptions, and choosing practical data-movement and validation approaches to reduce manual migration effort.

QixiaoWang introduces the Fabric Notebook Toolkit (FNTK) preview, a CLI that lets coding agents interact with live Microsoft Fabric notebooks—reading and editing cells, running code, and pulling outputs and Spark diagnostics—so iterative debugging and validation can happen directly from an agent-driven terminal workflow.

Laura Jiang announces Agentic Autofix for GitHub Advanced Security in Azure DevOps, which generates proposed fixes for both code scanning alerts and vulnerable dependencies. The update uses repository context plus custom instructions/skills to better match team standards, and adds triage signals like EPSS and dependency scope to help prioritize what to fix first.

The Visual Studio Code Team shares the 1.144 Insiders changelog, including updates to the experimental session composer, richer Mermaid diagram controls in the Markdown editor, improved OAuth token isolation for MCP servers, and an early feature for comparing multiple models without opening a folder.

Julien Dubois shares practical stories from three Java Champions using GitHub Copilot in real projects, focusing on agentic workflows, verification gates, and staying accountable for what ships—from large content migrations to debugging tricky concurrency issues on Raspberry Pi.

Waldek Mastykarz tests whether swapping a real GitHub repo for a local bare git remote changes AI coding agent behavior during evals, focusing on a rebase scenario where the agent must decide whether to push without being asked.

Allison announces generally available REST API endpoints for retrieving and updating GitHub Copilot model policies across enterprises, organizations, teams, and enterprise apps, making it easier to centrally control which Copilot models are available at each scope.

Allison announces that several AI models have been deprecated across GitHub Copilot experiences as of October 19, 2026, and lists the supported replacement models. The post also calls out what Copilot Enterprise admins may need to change in model policies so alternatives show up in Copilot Chat.

Satya Nadella shares Microsoft’s view that AI models and agents should be treated as insider risks, with security controls enforced outside the model. He highlights expanded Project Perception capabilities aimed at continuously assessing agent risk, adapting permissions, and containing threats across enterprise security layers.

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.

Kayla Cinnamon explains how to move an existing GitHub Copilot app agent session into VS Code so you can review the diff, keep the full conversation context, make manual edits, run tests locally, and push updates back to the same pull request.

Amir Jafari announces the deprecation of the Fabric data agent “custom tool” integration in Microsoft Foundry and explains how to migrate to the Fabric IQ (OneLake Catalog) tool, including portal steps and an SDK/REST approach using MCP endpoints and Entra-based authentication.

Gerald Versluis explains why existing .NET MAUI apps can crash right after the splash screen when built with Xcode 27, and walks through migrating to UIKit’s scene lifecycle—either manually or by using the .NET MAUI CLI “maui-xcode27-migration” agent skill with GitHub Copilot CLI.

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.

midesa announces general availability of advanced NL2SQL and advanced DAX generation for Microsoft Fabric data agents, focused on improving query correctness for enterprise schemas. The post explains the agent’s explore-then-generate approach, new creator configurations (schema descriptions and topics), and internal benchmark results showing higher accuracy and lower latency.

Hayete Gallot outlines why enterprise AI agents should be treated as “agentic insiders” and argues for continuous, real-time supervision using “Guardian Agents” and layered controls across identity, permissions, data, endpoints, network, cloud, and runtime.

Nadim Abdo outlines how organizations can treat AI agents as “agentic insiders” and reduce insider-risk-style failures using identity, least privilege, containment boundaries, and continuously updated risk signals across Microsoft’s security stack.

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.