The inability to take video evidence as being at all related to truth. Presidential candidates saying "I never said that, that video is a deepfake". If deepfakes are normalized enough, then people will always have sufficient reason to disbelieve actually-true video if they don't want to believe it. If the production and dissemination of deepfakes were criminal, it (a) wouldn't be normalized, and (b) people would think twice about making deepfake videos if punishments were criminal rather than civil slaps-on-the-wrist like libel is. And criminal means actual police resources dedicated to finding creators, whereas a libel suit has only the power of subpoena.

> The inability to take video evidence as being at all related to truth.

That has been true for 38 of the past 40 centuries. Somehow I suspect making it 39 out of 41 won't be that big of a problem, especially compared to getting people to not take video evidence as being at all related to truth.

(Edit: in case it wasn't obvious: you can't take video evidence as being at all related to truth if you don't have any video evidence to so take.)

People have been able to do this via video editing and audio manipulation for a while now and much better at that than state of the art deepfakes, and yet this isn't a problem despite the huge incentives you imply that would motivate very powerful, wealthy groups with the means to easily pull it off. Video evidence is still a thing, too.

Criminalizing making fake videos of any kind is extreme authoritarian behavior and libel is already punished by law everywhere I know. If you don't like how libel is punished in your jurisdiction that is another matter, but if you want to get some ideas on how to make them harsher just get a list of countries and filter by dictatorships and you're bound to find examples that take libel and its definition very seriously.