Recently, a netizen posted that a QR code covered in an overall blurred mosaic was fully restored by AI within a few minutes, and its content was scanned out directly. The poster also specifically reminded others that when masking a QR code, they should avoid redaction methods that support reversible operations.
As soon as the post appeared, reactions split into two camps: some worried that this kind of technology would amplify privacy leak risks, while others doubted whether "AI restoring a mosaicked QR code" was actually reliable. At present, the mosaic options offered by various social platforms and photo-editing software are diverse, with the three most commonly used being blur, brush smearing, and pixelation.
Of the three redaction methods, which is easiest to break?
To verify the claims, reporters from a media outlet created their own QR code and tested the three redaction methods separately. The results differed markedly: the QR code with overall blur was easily restored by AI; those with partial line masking and brush smearing could not be scanned from any of the images AI repeatedly generated; slight pixelation was easily cracked, while high-intensity pixelation could block restoration.
The pattern actually lies in the "amount of residual information" — the more recognizable clues left after redaction, the greater the room for AI to fill in the gaps. Blur and light pixelation happen to retain enough structural features, making reverse reconstruction nearly effortless.
The interviewed expert's conclusion was blunt: ordinary mosaics offer very poor protection, and whether a QR code can be restored depends on how much information remains after redaction. The safest approach is not to cover it, but to crop the QR code directly out of the image; if the image must be kept, it should be completely covered with an opaque solid-color block.