I think the biggest casualty of deepfakes are going to be citizens. Record a video of police brutality or crime footage and it can all be shrugged off as a deep fake. TV stations and journalists can afford 'signed videos'.
So you could easily take existing footage and change who it appears to be. But,
1) doing that to frame or unframe someone will be pretty niche 2) the original correct footage might still be findable 3) with access to the video files(which is going to be required in any legal situation), I'm sure there will be (exists for photos) algorytms that detect the editing the deepfake did to video file.
Of course, knowing all that is not really necessary for someone to shrug off truthful incriminating footage as “fake news”.
[0] One caveat is that deepfaking works better with higher quality footage. To produce a complicated scene with altered actor faces while maintaining realistic “phone video” look, it would make sense to film with good lighting and high-quality gear into log or raw format, and then imitate the look in post-production after deepfaking is applied.
Thus, one method of detecting such fakes could be by checking for traces of VFX, artificial noise, signature lens properties, signature behavior of phone video recording “magic” (such as noise reduction and stabilization), etc. Enough of producer’s dedication could make that tricky, but IMO it could be easier than applying automated deepfake detection straight up—it’d be buried early enough in post-production workflow, with a lot of noise introduced by subsequent “phone look” VFX.
This was treated (along with the ubiquity of CCTV in London) by the 2019 BBC show The Capture[0].
They glossed over a bunch of technical issues, but got the idea across pretty well.
Of course, they made the Americans look like the worse bad guys (no one comes out looking good, even the nominal "victim.") but that's to be expected, given the darkness the show attempts to purvey.
It's not great, but it absolutely provides context for the eventual advent of reasonable quality deep-fakes and the potential for abuse of the technology.
I'd expect that we'll see lots more of this sort of story telling, especially since any sort of video production (not including raw video footage) is never seen the way things actually are. As such, the "faking" (minus the deep-fakes with faces, etc.) is the key to the exercise.
Go watch the taping/filming of any television or movie and that will become immediately clear.
[0] https://en.wikipedia.org/wiki/The_Capture_(TV_series)
Edit: Added the missing link.
If someone has confirmation bias, they're still going to find ways to call it fake with or without deep fake tech.
People don't need deep fake videos to believe lies, they do it with little to no supporting evidence anyways.
I could easily see a deep fake viral video go around in a sub-Saharan country, with a leader claiming he is gay or something else socially unforgivable in that country, and starting an internal ethnic conflict that cannot be undone.