AI is shaking up the photography world in 2023, from fooling judges to win photography competitions to the proliferation of deepfakes with real-world consequences, such as an image of an explosion – that didn't happen – outside the Pentagon that caused the markets briefly collapsed.
Deepfakes can be funny (the Pope in a puffer jacket, anyone?), but they can also be used to spread disinformation, and as AI image generators become more powerful, it's becoming increasingly difficult to distinguish the real photos we see online from fakes .
Now it appears from a report from Nikkei AsiaLeading camera brands Sony, Canon and Nikon want to tackle this problem by building in technology that can verify the authenticity of images in new cameras.
Last year, the Leica M11-P became the first camera with built-in Content Credentials – a digital signature that verifies the time, date and location a photo was taken and who took it, and also indicates whether any edits were made after the recording.
However, not many people have an $8,000/£7,000 Leica in their bag, and now Sony, Canon and Nikon are about to introduce their own authentication technology.
We don't yet know which new cameras will hit shelves with built-in Content Credentials-type technology – although we've rounded up the 12 most exciting cameras for 2024, which might give us an idea. With the launch of the Sony A9 III, Sony announced that it will update that camera, and two other professional models, the A1 and A7S III, with support for Content Provenance and Authenticity (CP2A), although we don't know when or how. plan is to do this this. (CP2A is a cross-industry coalition co-founded by Adobe that introduced Content Credentials and includes Nikon and Getty.)
The Nikkei Asia report says future cameras will provide the necessary information to authenticate an image in a free and publicly available way Content Authenticity Initiative's (CAI) Verify system.
Analysis: Are digital signatures in the camera enough?
Increasing the number of cameras with authenticity technology that can mark images with the official Content Credentials stamp is a big step in the right direction, but is it enough?
Even as three of the biggest camera brands implement professional authenticity/anti-AI features, early indications are that these will be reserved primarily for professional cameras typically owned by journalists.
While major media organizations can implement protocols to fact-check and authenticate the provenance of images through the Content Credentials Verify tool for cameras like the M-11P, the majority of cameras will not be properly verified – including the ubiquitous cameras on smartphones from Apple, Samsung and Google, among others.
The bigger challenge, which these verification measures do not address, is websites and social media platforms that host disinformation, with fake images potentially seen and shared by millions of people.
I'm encouraged that camera brands are about to introduce digital signatures in future cameras and possibly update select existing cameras with this technology. But it appears it will be some time before these features are rolled out more widely to cameras and phones, including devices used by people looking to spread disinformation with fake images.
It's also unclear whether bad actors will find ways to circumvent digital signatures, whether they are AI-generated images or real images that have been edited. And what about sequences of multiple images in the camera, such as composites? Hopefully these questions will be answered as camera brands put these authentication measures into practice.
For now, the long battle against deepfakes has only just begun.