I don’t really know. But based on other risky situations I think maybe the certainty it is based on some variables to quantify it.
1.- The punishment that Twitch receives when that happens (legal or losing users)
2.- The cost to check that
3.- How the service is not actually live it they add controls.
There is also some predictability in users, then they should implement risk analysis depending on the user. Then if they have no expensive punishment they would prefer to control after they have been reported by another user.
It is also interesting that we demand “perfect” controls to virtual and IT products, but we do not demand that in our lives. e.g. What we see in the streets is “live” … then… Which warranty we have that none of the free people that walk in the street doesn’t make things that are inappropriate for children? What happens if an entire community do that, and the policeman (maybe) is not enough to prevent it?