Apple is sidestepping the photo-provenance standard. That could split how the internet proves what's real.

A photograph used to prove itself. The camera caught the light, the shutter clicked, the negative held the trace. Then generative AI made the trace reproducible, and the cameras started signing their work.

Apple is about to pick a different side of that fight.

Code references in the iOS 27 beta 5 build describe a feature called Apple Reference Image, a system that lets an iPhone embed provenance metadata into a photo at the moment of capture and later verify, on Apple's servers, that the file has not been silently rewritten. The Verge first reported the feature on top of 9to5Mac's original discovery, and MacRumors matched the detail against the iOS 27 beta privacy disclosure.

The mechanism looks a lot like something the camera industry has been putting in its cameras for the past few years. That something is C2PA Content Credentials, a provenance standard maintained by the Coalition for Content Provenance and Authenticity that signs a photo with a tamper-evident record of where it came from, what was done to it, and whether a generative model touched it. Canon, Nikon, Sony, Fujifilm, and Leica have introduced it; Google added it to the Pixel 10 earlier this year. Apple is not on that list, and Reference Image is the shape of its no.

The feature is off by default. To use it, a user has to flip on a toggle buried in Settings > Camera > Reference Image > Reference Mode, then take the photo with a new Reference option in the Camera app. Only files shot through that path carry the metadata required for authentication. Apple's beta privacy disclosure makes the trade-off explicit: tapping the Reference badge on a photo sends the raw image and its embedded provenance, including sensor signatures, a capture time window, and unique hardware identifiers, to Apple's Private Cloud Compute servers. Apple says the raw photo is not accessible to the company during verification. The sensor data is, and is sufficient for Apple to block images from a compromised sensor and to retroactively revoke authentication on photos the system has already signed.

The C2PA 2.4 specification distributes trust across the participating camera vendors and the software that re-signs edits. Apple's path concentrates it. A photo that Apple says came from an iPhone, Apple can also un-say.

The Verge frames Content Credentials as "not known for its reliability," a reference to known problems with stripped metadata, partial implementations, and a verification flow most readers never see. Apple's reasoning for a homegrown stack, on that reading, is a bet that one company running one closed verification pipeline will do better than a coalition standard that has been uneven in the field. Whether that bet holds is an open question; the beta is the first public place to check it against an actual Apple text rather than a leaked rumor.

iOS 27 has to ship with Reference Image enabled before any photographer can use it; the current switch lives in beta 5 and is not on by default. The first authenticated photos will come from people who already toggle new iOS settings on day one. The rest of the photo world, including the cameras that ship with Content Credentials baked in, will keep signing the C2PA way. Two provenance stacks, two verification paths, and a near-term chance that the same news photo carries two different authenticity stamps depending on who shot it.