Apple developing photo verification system to prove iPhone images aren't deepfakes

Apple is working on a feature called Apple Reference Image that aims to solve the problem of deepfakes and AI-manipulated images by letting iPhone users cryptographically prove their photos are authentic. According to code discovered in iOS 27 beta 5, the feature embeds provenance metadata directly into photographs at the moment they are captured using the iPhone camera, creating a digital record of when and how the photo was taken.

When enabled—via Settings > Camera > Reference Image > Reference Mode—the feature only works with photos taken using a new "Reference" option in the iPhone Camera app. These special Reference images carry embedded provenance information including sensor signatures, the precise capture time frame, and unique hardware identifiers that act as a fingerprint for the device that took the photo. To verify a photo, users tap a Reference badge on the image, which triggers Apple's authentication process. That process sends the raw image and its embedded provenance data to Apple's Private Cloud Compute servers, where the data is checked against Apple's records. The servers then return an authenticated version of the photo with a uniquely assigned ID, which can be displayed on iPhone, iPad, or Mac.

Apple's authentication system does not require the company to access the raw photo itself during verification—a privacy-focused design choice. However, the company may receive sensor data in a way that allows it to prevent images from compromised sensors from being authenticated, or to retroactively revoke authentication on images associated with a sensor that has been flagged.

The feature is not yet live and will be off by default when it eventually rolls out. Apple's approach differs notably from the C2PA Content Credentials standard, which major camera manufacturers—Canon, Nikon, Sony, FujiFilm, and Leica—have been gradually integrating into their hardware. Google's Pixel 10 phone cameras also support C2PA. Despite the industry convergence around C2PA, Apple has opted to build its own proprietary system, apparently believing it will be more robust. C2PA itself has not earned a strong reputation for reliability, which may have influenced Apple's decision to develop an alternative. According to reporting, Apple's focus on labeling and authenticating human-made content aligns with thinking from Instagram head Adam Mosseri, who has suggested that proving authenticity of human work may be more practical than identifying and labeling AI-generated content.