This video is mostly on the question of, “how do you police the internet”, he discussed 5 main methods, which all of them have flaws:
5:08: human moderation. Not fast enough with the rapid pace of things getting posted, and currently, there are already too many things to moderate.
7:37: automation (static), prone to false positive and negatives, and users can alter content to evade it.
9:07: automation (heuristic/learning). It takes forever and bad actors and feed it with wrong information which tricks it into allowing banned content.
13:31: top-down moderation. Nuclear option, too harsh.
22:26: laws. Very ineffective as criminals “go underground”, hiding from law enforcement. Criminals aren't stupid.
So, how do you police the internet, the answer is, you don't. It's as if gun manufactures are to be held liable when innocent people get murdered. Anything and services can get abused for illegal acts, and punishing the them isn't a good idea.