People don't underestimate it, they just don't know what to do about it. That's arguably more of a technological problem than a policy problem, the technical and philosophical problem of how to detect fakes.
Fake images on social media are much more than just a nuisance - they represent a fundamental threat to liberal democracy. Our brains simply aren't evolved to cope with convincing but fake images and we don't have any effective countermeasures.
The more promising angle is imho the society angle. Most people view their memory as the best source of truth, when in reality it's well established in psychology that our memory is incredibly unreliable. Previously this was mostly a problem for the justice system, now we risk entire nations being gaslighted. What we should be doing is trusting our written word, not our memory. Basically writing diaries and keeping newsletter articles, and checking both from time to time to keep wrong memories from manifesting
All of that is very teachable
Most of the issues discussed in the presidential elections are way beyond personal observation. Are we to throw our votes to the wind since it falls outside our experience?
Unless of course the attacker has vastly more computational power than the person trying to detect the fake
Throw billions of dollars at data science prizes to create open sourced ways of detecting deep fakes. Have western intelligence agencies keep some methods private too, so we can detect it when someone has a way of getting through the open sourced scanners. Figure out how then did it, then open source a third way of detecting what they did while keeping the private scanners private.
This is so obviously a problem to everything that we simply cannot allow this threat to be tolerated. Truth itself cannot be under attack. We simply must be able to trust video.