Microsoft has begun testing a new AI-powered right-click menu in Windows 11’s File Explorer, bringing one-click image edits and visual search directly to the context menu. The experiment, rolling out to Windows Insiders in the Canary Channel, marks a significant step toward embedding generative AI capabilities where users already interact with files. Alongside the new “AI actions” submenu, the preview build reintroduces a seconds clock in the Notification Center and adds a privacy page that reveals which apps have used Windows-provided generative AI models.

The changes were spotted in Windows 11 Insider Preview Build 27938, though community reports note that Canary features are heavily server-gated. Not every device on the same build will see the new options, and Microsoft has yet to confirm the feature set in its official Flight Hub. For now, the build number and feature details come from insider reports and the TechWorm review of the experimental release.

What’s new in this Canary experiment

The visible additions fall into three categories:

  • AI actions in File Explorer: A new submenu on right-click for supported image files, offering Bing Visual Search, Blur Background, Erase Objects, and Remove Background.
  • Notification Center seconds clock: An opt-in toggle under Settings → Time & language → Date & time that restores a larger clock with seconds display.
  • Generative AI privacy controls: A new panel under Settings → Privacy & security → Text and image generation that lists third-party apps that recently used Windows’ AI models, with per-app toggles to block access.

These changes reflect Microsoft’s strategy of making AI actionable at the file level while giving users and administrators visibility into generative AI usage.

AI actions in File Explorer: a deep dive

When the feature is active on a Canary build, right-clicking a .jpg, .jpeg, or .png file reveals an “AI actions” menu with four tools:

  • Bing Visual Search: Uses the image as a search query to find similar visuals, products, landmarks, or identify elements in the photo. Results are web-based, so an active internet connection is required.
  • Blur Background: Opens the Photos app and automatically applies a portrait-style blur, with sliders and brush tools for fine-tuning. It’s a quick way to create social-ready images without launching a full editor.
  • Erase Objects: Taps into Photos’ generative erase capability. Users can select unwanted elements, and the AI fills the removed area with plausible content. This mirrors tools like Google Magic Eraser but lives in the context menu.
  • Remove Background: Sends the image to Paint’s background removal pipeline, instantly cutting out the subject. Ideal for quick thumbnails, collages, or presentations.

Not all file types are supported: RAW, PSD, and TIFF files are not handled by these quick actions, and document-level tasks like summarization or FAQ generation are planned but not yet live. Microsoft has indicated that such features may require a Microsoft 365 Copilot license when they arrive.

How the flows execute

The implementation is hybrid. Some actions may run locally on devices with neural processing units (NPUs), like Copilot+ PCs, while others offload to the cloud. Microsoft has not published per-action locality guarantees, and community testers have observed inconsistent behavior. This ambiguity is a concern for privacy-conscious users handling sensitive images. Until official documentation clarifies data handling, assume that any action could involve cloud processing.

Notification Center clock makes a comeback

For users who missed the seconds display from earlier Windows versions, the Canary build adds a toggle: Settings → Time & language → Date & time → Show time in the Notification Center. When enabled, the clock in the system tray flyout shows hours, minutes, and seconds. It’s a small quality-of-life tweak that addresses a long-standing user request with no impact on performance.

Privacy visibility: who’s using Windows AI models?

A new page under Settings → Privacy & security → Text and image generation lists third-party apps that have recently accessed Windows-provided generative AI models. Each app has a toggle to allow or block future access. This is a transparency measure that lets users see which applications are tapping into AI capabilities like text generation or image manipulation.

For enterprises, the visibility is a starting point but falls short of full manageability. IT administrators will want policy-based controls that can restrict specific AI actions, cloud offload, or enforce audit logging. The current per-app toggles only give a binary on/off; there’s no granularity to allow an app to use local inference but block cloud processing, for example. Community discussions highlight the need for Group Policy and MDM integration before businesses can adopt these features at scale.

Performance and hardware considerations

Microsoft’s hybrid AI model means performance will vary. On Copilot+ hardware with dedicated NPUs, image edits like background blur or erase may execute quickly and locally. On older or lower-spec devices, the same action might ship the image to Microsoft’s cloud, adding latency and consuming bandwidth. Users should keep the Photos and Paint apps up to date via the Microsoft Store, because the File Explorer actions orchestrate through them, and version mismatches can cause failures.

Enterprise implications and licensing fragmentation

Businesses face several challenges with this AI rollout:

  • Licensing gates: Document-level AI actions are expected to require Copilot licenses, creating a rift between consumer and commercial experiences. Organizations with mixed licensing may see inconsistent file-action menus across employee devices.
  • Compliance risks: Without clear documentation on data flows and retention, enterprises cannot guarantee that sensitive files remain on-premises. The new privacy page offers visibility, but comprehensive audit logs and SIEM integration are necessary for regulated industries.
  • Manageability gaps: Early builds lack Group Policy support for AI actions. Administrators cannot yet enforce rules like “allow visual search but deny object erase” or mandate local-only processing. Microsoft has signaled that management tools will evolve, but timelines are vague.
  • Testing instability: Canary builds are unstable. Any deployment to production machines is strongly discouraged. IT shops should test on isolated devices and provide feedback through the Feedback Hub.

How to try these features now

To test the experimental AI actions and privacy controls, you must:

  1. Join the Windows Insider Program and enroll your device in the Canary Channel (expect rough edges and frequent updates).
  2. Update to the latest available preview build. Community reports point to Build 27938, but server-side gates mean not all Insiders will see the features.
  3. Right-click a supported image file (.jpg, .jpeg, .png) in File Explorer and look for AI actions. Try each option.
  4. Visit Settings → Time & language → Date & time and toggle on Show time in the Notification Center to enable the seconds clock.
  5. Check Settings → Privacy & security → Text and image generation to review app activity and manage toggles.

Caution: Some enthusiasts use ViVeTool to force-enable hidden features, but this is unsupported and can destabilize your system. Only attempt such tweaks on non-critical test machines.

Benefits: what Microsoft gets right

  • Reduced context switching: Simple edits no longer require opening an app, hunting for a tool, then switching back. Right-click, blur, done. Over a day of small tasks, the time savings add up.
  • Discoverability: Hiding AI behind a dedicated Copilot icon limits adoption. Placing actions right in the context menu—a familiar interaction—introduces generative AI capabilities to mainstream users organically.
  • Platform consistency: By routing actions through existing apps like Photos and Paint, Microsoft avoids reinventing the wheel and keeps the experience cohesive.

Risks and open questions

Despite the convenience, several concerns linger:

  • Privacy opacity: Without formal locality assurances, a hospital accidentally removing a background from a patient photo could unknowingly send the image to Microsoft’s servers. Microsoft must clarify exactly which actions run locally versus in the cloud and what data is transmitted.
  • Menu bloat: File Explorer’s context menu is already crowded. Power users may resent additional entries, especially if they cannot customize or hide them. Microsoft should offer a way to declutter or disable the AI actions submenu.
  • Artifacts and accuracy: Generative erase and background removal don’t always produce perfect results. Hair strands, complex patterns, and overlapping objects often yield artifacts. The one-click convenience may mislead users into trusting edits that require manual correction.
  • Feature fragmentation: When document-level AI actions roll out behind a Copilot paywall, the user experience will diverge. A consumer might see only image tools, while an enterprise user sees additional options—but only if their admin has paid for it. This could cause confusion and training overhead.

What this means for Windows’ AI future

Microsoft is clearly pushing toward an “invisible AI” model where intelligence is not a separate app but a utility woven into the OS shell. The File Explorer quicks actions are a prototype of that vision. If the company addresses privacy, manageability, and user-choice concerns, this approach could make AI genuinely useful for everyday tasks without demanding new workflows.

The next milestones to watch: official build confirmation on the Windows Insider blog, per-action locality documentation, Group Policy and MDM support, and the expansion of AI actions to document types and broader Insider rings. For now, adventurous users with test hardware can get a sneak peek at how Windows 11 might soon make AI editing as simple as a right-click.

Practical recommendations

  • For individuals: Test the features on a non-essential machine. If you deal with sensitive images, stick to local editing until Microsoft clarifies cloud processing. Check the privacy settings page afterward to see if any apps invoked AI models.
  • For power users: If you find the new menu cluttered, hold tight for customization options or document your feedback. Avoid registry hacks unless you’re prepared for instability.
  • For IT administrators: Start planning for AI governance. Inventory which file types your users handle, assess data sensitivity, and demand clarity from Microsoft on compliance controls. Don’t deploy Canary builds in production.

The experiment shows promise, but it’s a glimpse of a future still taking shape. Microsoft’s willingness to expose these raw features early in the Canary Channel suggests a commitment to community feedback—a practice that could make the final product more polished and privacy-aware.