Apple developing tech to verify photos shot on iPhone

References buried in the latest iOS 27 beta 5 build point to a new capability aimed at verifying the camera origin of iPhone photos. Here is what we know so far.

Apple is quietly developing a feature called Apple Reference Image, a system built to authenticate the source of a photograph using unique data tied to the iPhone camera hardware that captured it. The move comes as the problem of synthetic images masquerading as real ones grows more urgent, fueled by the rapid rise of generative AI.

The barrier to creating convincing fake imagery has collapsed. What once demanded hours of painstaking work from a skilled photo editor can now be produced in seconds with a loosely worded prompt. Recognizing this shift, several tech companies have rolled out tools to flag AI-generated content. Google, for example, created SynthID, a system that hides imperceptible watermarks inside AI-created images and media. That technology has since spread beyond Google’s ecosystem, with OpenAI among the adopters.

Meta has taken a different route, adding labels for AI-generated content across Facebook, Instagram, and Threads. Yet those efforts have struggled with reliability, sometimes failing to flag manipulated media even when produced by Meta’s own models.

Apple appears ready to enter the fray. Beyond the metadata it already attaches to images generated through Image Playground, new code in iOS 27 beta 5 reveals the company is building Apple Reference Image. The system is designed to authenticate photos using hardware-specific data from the iPhone camera that took them.

True to Apple’s privacy-first philosophy, the design prevents the company from ever seeing the raw photo itself. The image is sent to Private Cloud Compute for verification, while only select sensor information and photo metadata are transmitted back to Apple. Interestingly, if Apple determines a sensor may have been compromised, it can revoke earlier authentications tied to that sensor.

The feature is off by default. Once enabled, users will find it under Settings > Camera > Reference Image, with a toggle for Reference Mode. Activating it triggers a privacy splash screen, underscoring the sensitivity of the data involved.

A key limitation: Reference Mode must be active when the photo is taken. Standard shots taken without it will lack the provenance data required for later authentication. The mode itself does not automatically verify an image. Instead, users can choose to authenticate a photo after the fact, at which point the raw image and its provenance data are sent to Private Cloud Compute. During this process, Apple assigns the photo a unique ID.

Privacy remains a central pillar of the system. Authenticated photos can be shared with others, and their Apple devices can check locally whether the image is still verified, all without notifying Apple which photos are being viewed.

The data included when sharing depends on the method used, such as AirDrop or Messages, and whether the photo is an original, edited, or authenticated version. Notably, if a user opts to share “All Photos Data,” Apple warns that the transfer may include unique hardware identifiers and any uncropped footage tied to the image.

For USB transfers, Apple is adding a “Transfer with Provenance” option. When enabled, authenticated photos moved to a Mac or PC carry their Reference Image data. Unauthenticated photos may still include provenance details, such as hardware identifiers and uncropped footage.

What do you think of Apple Reference Image? Share your thoughts in the comments.

Worth checking out on Amazon