On Device Llm
The latest On Device Llm coverage — news, analysis, and updates from the WindowsNews.AI desk.
Microsoft Pays Anthropic to Add Claude Sonnet 4 as Copilot's Second AI Brain in Office 365
Microsoft is paying Anthropic to integrate Claude Sonnet 4 into Office 365 apps, adding a multi‑model orchestration layer to Copilot that routes tasks to the best‑fit backend—OpenAI, Anthropic, or in‑house models. The shift promises better performance on workloads like PowerPoint and Excel, cost savings, and vendor diversification, but introduces cross‑cloud complexity, data residency concerns, and the need for robust enterprise governance.
How Flyoobe 1.10 Lets You Install Windows 11 on Any PC and Ditch Copilot at Setup
Flyoobe 1.10 rebrands and expands the popular Flyby11 bypass tool into a full Out‑Of‑Box Experience customizer. It lets you install Windows 11 on unsupported hardware while disabling Copilot, removing bloatware, and scripting setup—all before first sign‑in. The article details how it works, its new features, and the real‑world risks including update uncertainty, security tradeoffs, and antivirus false positives.
Microsoft Ships SSMS 22 Preview 1 with Unified Settings, Theming, and AI-Driven IntelliSense
Microsoft has released SSMS 22 Preview 1, introducing unified settings, customizable themes, independent results grid zoom, and IntelliSense support for SQL Server 2025's AI and regex functions. The preview runs side-by-side with SSMS 21 but lacks SSIS and Copilot integration, with those features planned for later builds. Early adopters are urged to pilot the update in non-production environments to benefit from the productivity enhancements.
Microsoft Shuts Employee Forums and Mandates 3-Day Office Week in AI Blitz
Microsoft has imposed a sweeping workplace reset, shutting down open employee forums on Viva Engage, hardening campus security after a protest, and mandating a three-day office week for staff near its headquarters. The changes aim to accelerate AI product cycles but risk chilling internal dissent and driving talent away. The long-term success depends on transparent exception processes and measurable collaboration gains.
Windows 11 September Update Packs 4GB of AI Models You May Never Use
Microsoft's September 2025 Windows 11 update (KB5065426) introduces useful features like a Recall home page, expanded Settings AI agent, and Task Manager CPU fix, but buries a 3.8 GB payload of Copilot AI model binaries that every PC downloads—even without Copilot+ hardware. The shift signals a servicing model where on‑device AI components are shipped universally, raising bandwidth and storage concerns for users and IT administrators.
Ralph Lauren Launches AI Shopping Assistant Ask Ralph, Powered by Microsoft Azure OpenAI
Ralph Lauren has launched Ask Ralph, an AI-powered conversational stylist built on Microsoft Azure OpenAI, within its mobile app. Available initially in the U.S., the assistant provides brand-curated, shoppable outfit recommendations in a single conversational flow. While the tool showcases a strategic move toward first-party conversational commerce, its long-term success will depend on grounding accuracy, privacy safeguards and vendor independence.
Satya Nadella’s 5 Copilot Prompts Turn Microsoft 365 into an AI Leadership Engine
Satya Nadella published five practical Microsoft 365 Copilot prompts that leverage GPT-5 and Smart Mode to synthesize emails, meetings, and files into decision-ready briefs. The templates cover meeting prep, project updates, launch readiness probabilities, time audits, and email-anchored briefings, but require careful governance to manage privacy, provenance, and accuracy risks.
Visual Studio 2026 Debuts Agentic Copilot Superpowers—and New Attack Vectors
Microsoft released Visual Studio 2026 via a new Insiders Channel, deeply integrating agentic Copilot features powered by MCP and BYOM capabilities. While promising major productivity gains through adaptive AI coding, the launch has surfaced serious security risks exemplified by the 'Toxic Agent Flow' attack, demanding rigorous governance and least-privilege safeguards for safe enterprise adoption.
With $80B AI Bet, Microsoft Mandates Three-Day Office Return and Cracks Down on Internal Speech
Microsoft is rolling out a phased three-day-per-week return-to-office policy with a Redmond compliance target of February 2026, alongside tightened moderation of internal discussion platforms and enhanced campus security. The changes are framed as necessary to accelerate AI development amid an $80 billion infrastructure investment and in response to recent on-site protests. These moves raise concerns about talent retention, internal dissent, and the consistency of Microsoft's hybrid work narrative.
Microsoft Clamps Down on Internal Dissent, Mandates 3-Day Office Return by February 2026
Microsoft has shuttered internal employee forums, tightened campus security, and announced a phased return-to-office mandate requiring three days a week for workers within a 50-mile radius, starting with Redmond in February 2026. The moves aim to boost AI collaboration but risk chilling employee speech and triggering attrition.
Ralph Lauren's 'Ask Ralph' AI Stylist Uses Azure OpenAI to Build Shoppable Looks in Real Time
Ralph Lauren launched 'Ask Ralph,' a conversational AI stylist powered by Azure OpenAI, inside its U.S. mobile app on September 9, 2025. The assistant translates natural-language prompts into shoppable outfit laydowns, compressing discovery and checkout into a single flow. While the white-label approach preserves brand DNA and owns first-party data, success hinges on inventory accuracy, privacy controls, and vendor exit planning.
Microsoft Enforces Three-Day In-Office Mandate: Redmond Employees Must Return by February 2026
Microsoft will phase out broad work-from-home flexibility, mandating a three-day in-office week starting with its Puget Sound workforce by February 2026. The policy aims to accelerate AI collaboration but risks talent loss and equity issues if not implemented transparently.
Microsoft Mandates 3-Day Office Return, Silences Internal Forums in AI Pivot
Microsoft has enacted a phased three-day office mandate, closed internal employee forums, and tightened campus security following a protest, framing the moves as essential for accelerating its AI strategy. While some near-term operational benefits are plausible, the changes carry significant risks around talent retention, free speech, and equity. The policy’s ultimate success depends on transparent execution and measurable outcomes.