Browse Artificial Intelligence News (421)

cindywang explains how GitHub Copilot agents can modernize legacy Java and .NET code inside Docker Sandbox microVMs, keeping host filesystem paths consistent while avoiding risky Docker socket mounts and tightening egress controls during dependency upgrades.

Blanca Li summarizes April 2026 updates to Microsoft Foundry Reinforcement Fine-Tuning (RFT): global training for o4-mini across 13+ Azure regions, new GPT-4.1 family model graders, and practical best practices plus pitfalls to help teams design reliable graders and scale fine-tuning safely.

Allison announces `gh skill`, a new GitHub CLI command for discovering, installing, updating, and publishing portable “agent skills” for AI coding agents (including GitHub Copilot), with a focus on version pinning and supply-chain integrity.

Jim Harrer shares a curated list of 20 VS Live! Las Vegas 2026 sessions now available on the Visual Studio YouTube channel, spanning AI/Copilot topics, modern .NET and C#, Azure cloud-native development, GitHub Actions, and practical productivity and architecture guidance.

Microsoft Threat Intelligence and the Microsoft Defender Security Research Team break down a Sapphire Sleet macOS intrusion chain that relies on social engineering and user-initiated AppleScript execution, and provide Defender detections, KQL hunting queries, and IOCs to help security teams spot and stop similar activity.

Allison announces that Anthropic’s Claude Opus 4.7 is rolling out in GitHub Copilot, including where it’s available (IDEs, CLI, and agents), which plans get access, and what admins need to enable for Business and Enterprise tenants.

Microsoft Fabric Blog introduces the generally available OneLake MCP tools, showing how an AI agent can discover Fabric items, inspect schemas, map OneLake storage, and assess mirrored-database health through a single natural-language conversation (no code or portal clicks).

Microsoft Fabric Blog explains what customers are asking about Fabric Data Factory vs Azure Data Factory (ADF): when to migrate, how mature it is, and what new capabilities you gain—like Mirroring to OneLake, Copy Jobs with CDC, Dataflows Gen2, Copilot assistance, and managed Airflow/dbt execution.

Allison explains a new GitHub Copilot Cloud Agent (CCA) admin capability: enterprise admins can now enable the agent for selected organizations (including via organization custom properties), and manage the policy through the AI Controls page or new REST API endpoints.

Yun Jung Choi explains that Azure MCP tools are now built into Visual Studio 2022 via the Azure development workload, letting developers enable an Azure MCP Server inside GitHub Copilot Chat to provision resources, deploy apps, and troubleshoot Azure services without installing a separate extension.

Cassidy Williams interviews GitHub Staff Software Engineer Brittany Ellich about building a personal “command center” app, focusing on how GitHub Copilot CLI and agent-based workflows supported the process from planning through implementation.

Phillip Misner and Stephen Finnigan explain how incident response changes for AI systems: non-determinism and high-volume output shift triage, containment, telemetry needs, and remediation verification, while many IR fundamentals (ownership, escalation, and communication) still apply.

Harshada Hole introduces Visual Studio’s Debugger Agent guided workflow, which uses a live debugging session to help you reproduce bugs, validate hypotheses via breakpoints and call stacks, and iterate to a verified fix with less manual setup.

Fernando Vasconcellos outlines evergreen cloud cost optimization principles and explains how AI workloads change cost patterns, with practical guidance on visibility, governance, rightsizing, and continuous review—framed around managing and optimizing spend on Azure over time.

Reenu Saluja breaks down the main Azure hosting options for production AI agents and explains when to use each, with a deeper walkthrough of Microsoft Foundry Hosted Agents (deployment, lifecycle management, observability, scaling, and invocation patterns).

David Sanchez lays out a practical DevOps playbook for teams adopting AI coding agents (including GitHub Copilot Cloud Agent), focusing on readiness prerequisites, human–agent collaboration patterns, pipeline changes, governance, and security controls needed to keep quality and accountability intact as non-human contributors scale up.

Joseph Katsioloudes introduces Season 4 of GitHub’s Secure Code Game, a hands-on set of challenges where you exploit and fix vulnerabilities in an agentic AI assistant (ProdBot) to learn real-world AI-agent security risks like prompt-based tool misuse, memory poisoning, and sandbox escape.

stclarke shares a Microsoft AI announcement introducing MAI-Image-2-Efficient, a production-oriented text-to-image model available in Microsoft Foundry and MAI Playground, positioned as faster and cheaper than MAI-Image-2 while maintaining “flagship” quality.

Ismael Mejía Useche and Pooja Yarabothu introduce the “PostgreSQL Like a Pro” video series, focused on practical ways to run and modernize PostgreSQL workloads on Azure—covering AI agent patterns, AI-assisted migrations in VS Code, and performance/resiliency considerations at scale.

Allison announces that GitHub Copilot’s third-party coding agents (Claude and Codex) now support model selection on github.com, including the currently available Claude Sonnet/Opus and GPT-5 Codex model options, plus notes on required admin policies and repo settings.