This line of argument seems uncomfortably similar to "we need to ban encryption because terrorists and pedophiles can use it". The fact that bad people can occasionally benefit from something isn't a reason to not give it to the rest of us.
It actually can be. There's a continuum.
At one extreme are things like plans for nuclear or biological weapons. Nobody distributes those because terrorists really want them for horrible ends and they're of low utility, so we do without them.
At the other extreme are things like junk mail - we all just put up with that minor abuse of the system because having a cheap, flexible mail system is way better than not having junk mail.
Like I said, it's a spectrum, and sometimes bad people being able to abuse something is a good reason for everyone to forgo it.
My understanding is that how to build nuclear weapons is relatively well-known, and the only reason that terrorist groups don't have them is that they can't get their hands on the required raw materials.
In practice we want to pick the course of action that minimizes harm. Having your video demonetized is one kind of harm, having videos of harassment or revenge porn posted online is another.
The kind of people who say "we should never have false positives, ever" are never the people who will have to defend these positions once somebody gets harmed by them and kicks up a scandal.
That sounds great until the scandals of the type "Someone reposted SFW segments of this girl's revenge porn video and Facebook is taking days to remove them every time; since they're reposted every 24h they stay up forever" roll in.
Every so often we get another scandal on HN of the type "This horribly harmful content was posted on PLATFORM and PLATFORM didn't remove it fast enough". Augmenting the false-negative rate would make the problem worse.
Attackers can use automation to post their harmful content. There's no way defenders can keep up unless they use automation as well, no matter how decentralized they are.