Apple Reference Image authenticates iPhone photos to fight deepfakes

Apple is seemingly developing a new iOS feature called 'Apple Reference Image' that could let users prove their iPhone photos aren't deepfakes. Discovered in iOS 27 beta 5 by 9to5Mac, the system embeds provenance metadata into photographs at the point of capture. Apple hasn't enabled it yet, but code suggests it will be off by default and can be turned on via Settings > Camera > Reference Image > Reference Mode. Only photos taken with a new 'Reference' option in the Camera app will include the authentication data. When a user taps a Reference badge on a photo, the raw image and its embedded metadata—including sensor signatures, capture time frame, and unique hardware identifiers—are sent to Apple's Private Cloud Compute servers. Apple verifies the data and returns an authenticated version with a uniquely assigned ID, without Apple accessing the raw image itself. However, Apple may receive sensor data to block compromised sensors or retroactively revoke prior authentication for affected images.

The system works similarly to the C2PA Content Credentials standard, which traces how and where an image was made and whether generative AI was used. Camera makers like Canon, Nikon, Sony, and Google's Pixel 10 support C2PA, but Apple has notably avoided adopting it—perhaps because C2PA isn't known for reliability. Apple's own system could prove more robust. The move also suggests Apple is throwing support behind labeling human-made works online, something Instagram head Adam Mosseri believes may be easier than identifying AI-generated content. Authenticated photos can be viewed on iPhone, iPad, and Mac, according to MacRumors. While the feature isn't live yet, Apple Reference Image represents a significant step toward trustworthy visual media at the device level, giving journalists, creators, and professionals a more dependable way to authenticate imagery in an era of increasingly convincing deepfakes.