I don't think the issue here is related to AI. Without AI, moderators would still have to look at these same videos. The difference is they would hit the public first before being flagged and sent to moderators. Now with AI they can be prevented from ever going public.

The fact that we still need to traumatize workers to confirm the automated decisions is sad. The only other ways I can see to resolve this would be either to just blindly trust the AI result without any human oversight, or to require all facebook users to link their government ID to accounts and only allow posting by users in countries where the authorities arrest the people posting these things.