How it works
MediaChecker SafeGuard is a customer-facing evidence review for selected media. It is built to make the reasoning easier to inspect, not to create a mysterious real” or fake” stamp.
The problem
Digital media can be resized, recompressed, captured through a screen, edited, generated, or shared without useful context. A single visual clue rarely tells the whole story. A confident-looking number can be less helpful than a careful explanation of what was actually observed.
What you get
For a piece of media you choose, SafeGuard provides a plain-language result built from several evidence paths. It keeps soft observations separate from stronger origin signals, records stage-level audits, and can say insufficient evidence when the file does not support more.
What is running for you
1. Select media
You choose the media to examine in the page or extension workflow. The beta works on the selected item and the context needed to identify it.
2. Examine texture and compression
The service looks for image-level patterns such as compression and ELA-style texture. These are clues, not proof of how a file was made.
3. Run a soft-capture / shaking-hand check
A separate path considers signs associated with a screen, glass, bloom, motion, or a shaking hand. This helps distinguish capture conditions from origin claims; a soft-capture clue is not automatically an AI or CGI finding.
4. Compare parallel origin scores
The system keeps possible origins side by side, such as camera, game, CGI, AI generation, or composite.