For decades, Microsoft Paint has been the digital equivalent of a crayon: simple, harmless, and local. But a recent discovery by researcher Xusheng Li has revealed that Paint (and the Photos app) is now doing something far more sophisticated—and stealthy. It turns out your locally generated AI images aren't as private as you think.
The Invisible Fingerprint
According to Li's reverse engineering, Microsoft is embedding server-issued GUIDs (Globally Unique Identifiers) as invisible watermarks directly into the pixels of AI-generated images. While the images look normal to the human eye, they carry a hidden digital signature.
Crucially, these watermarks aren't just marking the image as "AI-generated" for safety; they are linked to server-issued IDs. This means that even if you think you're creating art locally on a Copilot+ PC, there is a traceable link connecting that specific file back to a unique identifier.
Privacy vs. Provenance
Microsoft's likely justification is provenance—the need to track AI-generated content to fight misinformation via C2PA manifests. However, the community reaction has been far less forgiving. On platforms like Hacker News, users have pointed out that linking these marks to GUIDs effectively turns a creative tool into a device-tracking mechanism.
If a GUID can be traced back to a user account or a specific machine, the "local" nature of the generation becomes a moot point. Your art becomes a beacon, signaling exactly who made it and where it came from, without your explicit consent or knowledge.
What Now?
As AI becomes deeply integrated into the OS, the line between "local utility" and "cloud service" is blurring. This discovery serves as a stark reminder: in the modern Windows ecosystem, "local" rarely means "private."
Sources
Media
