Windows Insiders on the bleeding-edge Canary channel are discovering a new context menu that puts generative AI image edits—including background removal, object erasing, and Bing Visual Search—just two clicks away from any JPG or PNG. The feature, tied to Build 27938 in community reports, marks Microsoft’s boldest move yet to weave AI into the operating system’s core file management experience. Rather than forcing users to launch separate applications, these “AI actions” sit directly in File Explorer’s right-click menu, promising to slash the time it takes to perform common micro-edits. But beneath the surface of this slick integration lie unresolved questions about data privacy, cloud processing, and administrative control that temper the excitement.

Why File Explorer Is the New Battleground for AI

File Explorer has long been the nerve center for Windows users—the place where files are discovered, organized, and dispatched to apps. Over the past two years, Microsoft has layered in tabbed browsing, a redesigned Home view, and deeper OneDrive integration, all aimed at reducing context switches. Adding AI capabilities at the file level is the logical next step. By embedding assistive features where users already interact with their content, the company hopes to make routine tasks instantaneous. This strategy aligns with a broader OS-wide push: Copilot integration, on-device model surfaces on Copilot+ hardware, and now contextual actions that turn file management into an intelligent workflow.

What the AI Actions Menu Actually Offers

When the feature is active, right-clicking a supported image reveals an “AI actions” submenu with up to four options: Bing Visual Search (uses the image as a search query), Blur Background (launches Photos with automatic subject detection and adjustable intensity), Erase Objects (invokes Photos’ generative erase), and Remove Background (routes to Paint for a quick subject cutout). Initially, only JPEG and PNG files are supported, but Microsoft may expand format coverage based on insider feedback.

Build 27938 is the number circulating in insider forums, but Canary flights are notoriously server-gated. Two users on the same build may see completely different feature sets, so the build number should be treated as provisional until official release notes confirm it. Regardless, the menu has appeared for enough testers to validate its design intent.

Under the Hood: Shell Hooks, App Backends, and the Local-vs-Cloud Mystery

Technically, the AI actions entry is a shell hook—a context-menu launcher that either calls a first-party app (Photos or Paint) with a scripted edit flow, or invokes Windows’ generative AI platform APIs directly. In practice, most operations open the respective app with the effect pre-staged, meaning the installed app versions and platform APIs on your device dictate the experience. On Copilot+ certified hardware, some processing may run locally using an NPU; otherwise, it falls back to cloud endpoints.

Microsoft has not yet published a clear decision tree explaining which actions run locally versus in the cloud. This transparency gap is significant. Bing Visual Search almost certainly uploads images to Microsoft servers. Background blur and object erase have historically been cloud-reliant in Photos, though local models are emerging. For enterprises and privacy-conscious users, this ambiguity is a dealbreaker until clarified.

Practical Productivity Wins

For individuals who handle a high volume of screenshots, product images, or social media content, these shortcuts are genuine time-savers. Imagine removing a distracting background from a product shot without ever launching a separate editor, or quickly erasing a license plate from a screenshot before sharing it. Bing Visual Search can turn any image into a research launchpad, identifying products, landmarks, or similar images instantly. These micro-optimizations may shave only seconds per task, but they accumulate over a workday and keep users in flow.

How to Try It Right Now

To test the AI actions menu, you’ll need to join the Windows Insider Program and opt into the Canary channel (though Dev or Beta may receive it later). Update to the latest preview build, then right-click a JPEG or PNG in File Explorer. If the AI actions submenu doesn’t appear, remember that rollouts are staged. Power users often resort to tools like ViveTool to force-enable feature flags, but this can destabilize your system and is only recommended on disposable test machines.

Privacy, Security, and Governance: The Hard Questions

With any feature that touches cloud services, data handling becomes paramount. Bing Visual Search and certain Copilot flows will upload image payloads to Microsoft’s servers. Even if local models exist, the hybrid architecture means network activity can occur without explicit user consent—at least, that’s the fear among IT admins. Microsoft has added a new “Text and image generation” view under Windows’ Privacy & security settings, showing which apps have recently used generative models and offering per‑app toggles. That’s a welcome transparency step, but it’s only a start.

Enterprise administrators need more: per-action blocking, network restrictions, audit logging, and Group Policy/MDM controls. Today, those are largely absent. The attack surface also raises eyebrows—any mechanism that simplifies sending file contents off-device could be exploited by malware or a malicious insider to exfiltrate sensitive data. Until Microsoft delivers fine-grained IT controls, many organizations will understandably keep AI actions disabled or restrict Insider devices to non-production roles.

The UX Tradeoff: Discoverability vs. Clutter

Adding a submenu to the already crowded right-click context menu is a double-edged sword. Casual users will appreciate the discoverability, but power users who rely on muscle memory and keyboard shortcuts may find the added entries irritating. Without the ability to hide or reorder these items, the feature risks cluttering an interface that many consider sacred ground. Early insider feedback reflects this tension: convenience for some, annoyance for others.

Verification: What’s Confirmed and What’s Still Murky

  • Confirmed: Microsoft is actively testing AI actions that include Bing Visual Search, background blur, object erase, and background removal in File Explorer, as reported by multiple outlets and visible in insider previews.
  • Confirmed: Initial format support covers only JPG/JPEG and PNG.
  • Confirmed: Microsoft plans similar AI actions for Microsoft 365 documents (summarize, FAQ creation), but those will initially require commercial Copilot licensing before expanding to consumers.
  • Partially confirmed: The experiment is tied to Build 27938, but since Canary rollouts are server-gated, the build number should be treated as provisional until official notes appear.
  • Unverified: The exact split between local and cloud execution per action remains undocumented. Microsoft’s hybrid approach suggests device hardware and licensing will influence processing, but specifics are speculative.

Recommendations for Power Users and IT Teams

For power users and creators:
- Test these features only on a non-critical machine or a VM to avoid Canary instability.
- Keep the Photos and Paint apps updated via the Microsoft Store—their versions directly impact the AI actions experience.
- Avoid using Visual Search on images containing sensitive or confidential information until Microsoft clarifies data retention policies.

For IT and security teams:
- Inventory all Insider devices and ensure they are separated from production workloads.
- Evaluate the new generative AI settings surfaces and begin planning MDM/Group Policy configurations as they mature.
- Enforce network egress rules to restrict unintended cloud uploads from Photos and Paint processes.
- Update acceptable-use policies to explicitly address the new context-menu options and their data implications.

Broader Implications and What to Watch Next

Expect Microsoft to expand the action set—adding more image operations, supporting additional file formats, and integrating document summarization—as Insider feedback rolls in. Multiple reports point to a roadmap that includes AI actions for Office files and deeper Copilot integration. The company will also face pressure to document local versus cloud processing with absolute clarity; without it, enterprise adoption will stall. The initial Settings surface for tracking generative AI activity is a promising foundation, but enterprise-grade audit capabilities are still missing.

Design Critique: Strengths and Notable Risks

Strengths:
- Reduced friction: Small edits become file-level operations, eliminating context switches.
- Leverages existing apps: By routing actions through Photos and Paint, Microsoft accelerates rollout while building on proven backends.
- Transparency efforts: The new “recent activity” screen in Settings shows a commitment to visibility, even if it’s incomplete.

Risks and caveats:
- Privacy ambiguity: Unclear data flows could erode trust, especially for business users handling sensitive images.
- Enterprise governance gaps: Administrators need per-action policy enforcement and audit trails before deployment.
- Context-menu clutter: Without customization, power users may resent the added entries.

A Pragmatic, Cautious Step Forward

Embedding AI actions into File Explorer is a logical next step for Windows, meeting users where their files live and solving real, repetitive problems with minimal friction. For individuals who frequently edit screenshots or product images, the shortcuts will feel like a genuine productivity breakthrough. For IT pros and privacy-minded users, however, the preview raises important governance questions that Microsoft must answer before broad deployment. Treat the current build as a preview of Windows’ trajectory rather than a finished product. Test on non-production devices, follow the official Windows Insider blog for confirmed details, and watch closely as the privacy and administrative controls evolve. The experiment proves one thing: as AI becomes a routine part of daily computing, the operating system itself will increasingly become the surface for intelligent micro-workflows—and that demands a new level of transparency.