Security
Apple Photos Now Adds Invisible AI Watermarks to Edited Images
iOS 27.2 embeds hidden SynthID markers when users apply Clean Up, Extend, and other Apple Intelligence editing tools.
Key takeaways
- Apple Photos in iOS 27.2 adds invisible SynthID watermarks when users edit images with AI tools like Clean Up, Extend, and Reframe
- The watermarks are undetectable to the eye but can be identified using Google's SynthID Detector at synthid.com
- The system cannot distinguish between minor edits (removing a background person) and entirely AI-generated images
- Only photos edited with Apple Intelligence tools after the October 26 update receive watermarks; older photos remain unmarked unless re-edited
- Apple Intelligence editing features require iPhone 15 Pro models or iPhone 16 and newer compatible devices
Apple has begun embedding invisible watermarks in photos edited with its artificial intelligence tools, a change that arrived with the iOS 27.2 update released October 26.
The watermarks use Google's SynthID technology to mark images that have been created or modified using AI. Users won't see any visible indication on their photos, and recipients who receive shared images typically won't notice the marker either. However, the watermark can be detected using compatible scanning tools.
Why it matters
As AI photo editing becomes indistinguishable from reality, the line between authentic and manipulated imagery blurs. Apple's watermarking approach treats minor edits—like removing a stranger from a vacation photo—the same as entirely AI-generated scenes, creating ambiguity about what the marker actually signals. For business communications, journalism, and legal contexts where image authenticity matters, understanding these limitations is critical.
Which editing tools trigger the watermark
According to Apple's support documentation, several Apple Intelligence features now apply SynthID watermarks:
Clean Up removes unwanted objects or people from photos, with AI filling in the missing areas. Photos edited this way may receive a watermark, though Apple hasn't specified whether every minor edit triggers the marker.
Extend generates additional background beyond a photo's original edges, creating scenery that the camera never captured. Because AI creates part of the final image, these photos can carry the hidden marker.
Reframe (Spatial Reframing) changes a picture's perspective after capture, with Apple Intelligence generating necessary visual information. These altered images may also be watermarked.
Image Playground and Image Wand create images from descriptions or sketches. Since these tools generate content entirely through AI, their inclusion in the watermarking system is straightforward.
The feature applies to devices running iOS 27.2, iPadOS 27.2, macOS 27.2, and visionOS 27.2. Apple Intelligence editing features require an iPhone 15 Pro, iPhone 15 Pro Max, or iPhone 16 models and newer compatible devices.
How to detect the watermark
Users can check whether a photo contains a SynthID watermark by uploading it to Google's SynthID Detector at synthid.com. The online tool analyzes images and reports whether the marker is present.
Importantly, finding a watermark only confirms AI was involved in creating or modifying an image—it doesn't indicate the image is fake. Conversely, the absence of a watermark doesn't prove authenticity, since other AI tools may not use SynthID, and detection has limitations.
The authenticity challenge
The watermarking system faces a significant limitation: it cannot distinguish between minor edits and substantial fabrications. A family photo with one background person removed receives the same marker as a completely AI-generated scene that never occurred.
Photos created or edited before the iOS 27.2 update won't automatically receive watermarks. However, opening and editing an older photo with supported AI tools today will add the marker to the newly edited version.
These details were first reported by Fox News in coverage by Kurt "CyberGuy" Knutsson.