Why would any kind of camera photo signing algorithm fare better?
Reporters attesting to their newsroom the authenticity of a photo, libraries and museums having stronger chain of custody and documentation of the state of their collections, etc.
As far as hardware signing, perhaps PKI and per-camera keys could be issued. HDCP was encryption with global symmetric keys, no? This would be a different case.
You probably wouldn't even have to bother. If you want to fake history, just launder your images: setup rig where an SLR camera can telecine the images off a high resolution screen. With some optics you could probably also adjust the other camera parameters (like focus, f-stop) to make it more convincing.
Both those things can be faked. IIRC, they make chambers right now that simulate GPS signals for testing receivers. If you can fake GPS, you can fake the time, too.
I'm reminded of Facebook's progression over the past 5-6 years, where, in its established markets at least, most of the good stuff happens in private Facebook groups or private WhatsApp chats, not on the 'public' newsfeed/wall.
I don't argue that it will be impossible in the long term. Right now, though, AI doesn't seem to be able to solve fundamental new problems - only to see patterns that humans can't yet see. I don't know if it is yet expected for AI to do fundamental new research on algorithms.
I also don't think the person above you suggested it's not possible - but their point stands, AI is going to really fuck society up if there is literally no way to communicate further than seeing/hearing distance between humans without knowing if the message is genuine. We'll have solved cancer and reverted to the 1700s.