Buried in the code of Apple’s iOS 27 beta 5 is a reference to something called Apple Reference Image. The system could let iPhone owners prove that a photo came from their phone’s camera, not
Buried in the code of Apple’s iOS 27 beta 5 is a reference to something called Apple Reference Image. The system could let iPhone owners prove that a photo came from their phone’s camera, not an AI image generator.
The references surfaced on August 10; however, this feature is turned off in the current beta.
The tech industry has been tagging synthetic content for two years. Google’s SynthID embeds invisible watermarks in AI-generated pictures. OpenAI has also adopted the same technology for its AI-generated pictures. Meta is adding AI tags to Facebook, Instagram, and Threads.
Apple also watermarks images it generates, including output from Image Playground. The Reference Image feature links a photo back to the physical camera sensor that took it. That provides proof that a real lens made the frame.
The feature is disabled by default. A user would go to Settings, Camera, and Reference Image and toggle on Reference Mode. Authentication only works for shots taken in a dedicated Reference Mode in the Camera app. Pictures taken in the usual way have no provenance data.
In Reference Mode, the owner taps a Reference badge after capturing a picture. That sends sensor signatures, the capture time frame, and the sensor’s unique hardware identifiers to Apple’s Private Cloud Compute servers.
The servers judge whether the sensor took the picture, assign a unique ID to the photo, and send back an authenticated copy to the device.
iOS 27 beta 5 shows that Apple is working on “Apple Reference Image”, a new Camera feature for authenticating the source of photos.
When enabled, Reference mode embeds provenance information that can be used to verify the camera sensor that captured a photo. Authentication is… pic.twitter.com/M03XmfQgxb
— Aaron (@aaronp613) August 11, 2026
Apple designed the process so that its servers never see the raw photo. During verification, the company receives sensor data, a hash, and the assessment result. Only selected sensor information and metadata are sent to Private Cloud Compute.
Apple can refuse to verify photos from a sensor it believes has been tampered with. It can retroactively invalidate a sensor’s previous authentications.
An authenticated photo shared with another Apple device can be verified locally, without Apple knowing which pictures someone is looking at.
Apple is adding a “Transfer with Provenance” option for USB exports to a Mac or PC. A broader “All Photos Data” share may include hardware identifiers and any uncropped footage.
Mentions of “photos or videos” and “uncropped footage” hint at video authentication possibly coming next. The concept is similar to C2PA Content Credentials. That standard is already used by Leica, Sony, and Nikon, and Google adopted it for the Pixel 10 line.
The code is in an active beta, and the Reference Image feature may not make the cut when iOS 27 ships to the public this fall.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.
