There is a "relatively simple"[0] solution to this issue of deepfakes. Anyone publishing content, particularly if there is incentive for manipulation, can get a

There is a "relatively simple"[0] solution to this issue of deepfakes. Anyone publishing content, particularly if there is incentive for manipulation, can get a Keyoxide[1] or other similar doip setup. Then they also sign and provider a hash of any content they publish with the associated key. There's a bit of challenge for binary content since those can't be compressed or optimized without breaking the hash, and many hosting platforms will make such changes. For that we need to advance perceptual hashing algos to the point where they only - and always - break on semantic differences. [0] Caveat being getting the solution to be seamless enough for the not-so-tech-inclined to have little problem using it. And for the general public to get into the habit of verifying content. [1] https://keyoxide.org/ https://keyoxide.org/