Apple now hides a Google-made AI watermark in photos edited with Apple Intelligence

Edit a photo with Apple’s AI tools and the result now carries a hidden mark that says so. Since iOS 27.2, Apple Intelligence stamps an invisible watermark on images it creates or changes, so a cleaned-up holiday shot can later be recognised as AI-edited.

The twist is whose technology it is. Apple did not build its own system: the mark is SynthID, the watermark developed by Google DeepMind, and checking it means going to Google’s synthid.com and uploading the picture to its detector.

According to a support note Apple published on November 5, the watermark is added by five tools. In Photos those are Clean Up, Extend and Spatial Reframing; outside it, Image Playground and Image Wand. It comes with the 27.2 releases for iPhone, iPad, Mac and Vision Pro; the iPhone update began rolling out last week.

Apple says the mark cannot be seen and does not change how the image looks. Only a SynthID detector can read it. Pictures made or edited before the update stay unmarked until they pass through one of those tools again.

That makes even small fixes count: removing a stranger from the background with Clean Up is enough to tag a real photo as AI-touched, the same as a picture generated from scratch in Image Playground. Ordinary edits such as crops and filters are not on Apple’s list.

What the note does not mention is any switch to turn the watermark off, or whether Photos will ever show a visible label of its own.

Reported by Apple, 9to5Mac, Google DeepMind. Compiled with AI tools.