I disagree that "deepfakes" will be unverifiable. There is very robust forensic framework for analyzing videos for authenticity, and it often relies not just on the visual component but all the other vectors that open up as a result-EM waves being generated, audio analysis, etc. A deepfake that passes this analytical function would imply that we will not be seeing advancements in digital forensics due to AI, which does not make sense because there is no reason not to update forensic processes to keep up with changing technology.