Which iPhones support Apple Reference Image?

Apple Reference Image runs on iPhone 18 Pro and iPhone 18 Pro Max, and only on the main camera sensor. Apple says the mode depends on a custom-designed sensor that can sign pixel data at readout, so it is not a software feature that older iPhones can receive in an update. It is opt-in, meaning normal photos are unaffected unless the mode is switched on.

Can Apple Reference Image tell whether a photo was made by AI?

No — Apple Reference Image does not analyse a picture to guess whether a generator made it. It works the other way round: it gives a photo that came from an iPhone 18 Pro sensor a cryptographic proof of that origin. An image with no valid reference is simply unproven, which is not the same as proven fake, and Apple states the system cannot verify images from non-Apple devices.

How is Apple Reference Image different from C2PA?

C2PA attaches provenance metadata after a picture is captured, so Apple argues it can be compromised at any point in the editing chain. Apple Reference Image moves the signature down into the sensor at the moment of capture, before any software touches the pixels. Apple also claims it is the only image provenance system offering quantum-secure defenses, via its composite RSA-3072 and ML-DSA-87 signature.

Can anyone check that a reference image is genuine?

Apple Reference Image is verified by the client every time the picture is shown, and the signature chain is designed to be checkable rather than trusted on Apple's word. Experts can inspect the Private Cloud Compute software that develops the negative, because production builds are published in an append-only, tamper-proof transparency log. Apple's post does not describe a consumer-facing verification website.

What does Apple Reference Image not protect against?

Apple Reference Image does not claim to stop every hardware attack; it resists them and falls back on revoking a compromised signing key. It also cannot vouch for images from non-Apple devices, and it will not retroactively protect a picture that was already edited before the mode was used. The proof covers the path from sensor to finished reference image, not what happens to copies afterwards.

Apple · 2026-09-15 · major

Apple Reference Image — iPhone 18 Pro signs a photo at the sensor

Apple Reference Image is an opt-in camera mode on iPhone 18 Pro that has the image sensor cryptographically sign pixel data at the moment of capture, so a photo can later be shown to be a real camera capture.

An opt-in iPhone 18 Pro camera mode that proves a photo came from a real sensor, not from a generator.

Key specs

Signature scheme	RSA-3072 + ML-DSA-87

Quick facts

Maker	Apple
Devices	iPhone 18 Pro and 18 Pro Max (main camera)
How to turn on	Opt-in camera mode
Signature	Composite RSA-3072 + ML-DSA-87
Processing	Private Cloud Compute
Where you see it	Photos app, next to the original

What is it?

Apple Reference Image adds a new camera mode that produces a second, unalterable copy of every shot — what Apple calls a secure digital negative. The Photos app shows it beside the normal picture, so a viewer can compare the two and see whether the visible photo was edited. Apple's security research team published the design on 15 September 2026. The mode is opt-in and runs only on the main camera sensor of iPhone 18 Pro and iPhone 18 Pro Max.

How does it work?

The sensor itself signs the pixel data the instant it is read out, and the sensor firmware is blocked from changing that data first. Apple then sends the signed negative to Private Cloud Compute, which performs the normal development steps — demosaicing, tone mapping and compression — inside an environment whose production builds are recorded in an append-only transparency log and whose binaries can be inspected. The finished reference image carries a composite post-quantum signature and timestamp bounds from an Apple service, and the client re-checks that signature every time the image is displayed.

Why does it matter?

Most work on synthetic media tries to label what a generator produced; this inverts the problem and proves what a camera actually saw. Apple's own framing points at the gap: widely available AI tools now generate or alter photorealistic images easily. For newsrooms and courts, a sensor-level signature is harder to strip than C2PA metadata, which is attached after capture and can be broken anywhere in the editing chain.

Who is it for?

photojournalists, newsrooms and anyone verifying images

Frequently asked questions

Which iPhones support Apple Reference Image?: Apple Reference Image runs on iPhone 18 Pro and iPhone 18 Pro Max, and only on the main camera sensor. Apple says the mode depends on a custom-designed sensor that can sign pixel data at readout, so it is not a software feature that older iPhones can receive in an update. It is opt-in, meaning normal photos are unaffected unless the mode is switched on.
Can Apple Reference Image tell whether a photo was made by AI?: No — Apple Reference Image does not analyse a picture to guess whether a generator made it. It works the other way round: it gives a photo that came from an iPhone 18 Pro sensor a cryptographic proof of that origin. An image with no valid reference is simply unproven, which is not the same as proven fake, and Apple states the system cannot verify images from non-Apple devices.
How is Apple Reference Image different from C2PA?: C2PA attaches provenance metadata after a picture is captured, so Apple argues it can be compromised at any point in the editing chain. Apple Reference Image moves the signature down into the sensor at the moment of capture, before any software touches the pixels. Apple also claims it is the only image provenance system offering quantum-secure defenses, via its composite RSA-3072 and ML-DSA-87 signature.
Can anyone check that a reference image is genuine?: Apple Reference Image is verified by the client every time the picture is shown, and the signature chain is designed to be checkable rather than trusted on Apple's word. Experts can inspect the Private Cloud Compute software that develops the negative, because production builds are published in an append-only, tamper-proof transparency log. Apple's post does not describe a consumer-facing verification website.
What does Apple Reference Image not protect against?: Apple Reference Image does not claim to stop every hardware attack; it resists them and falls back on revoking a compromised signing key. It also cannot vouch for images from non-Apple devices, and it will not retroactively protect a picture that was already edited before the mode was used. The proof covers the path from sensor to finished reference image, not what happens to copies afterwards.