https://www.wnycstudios.org/story/post-no-evil
From what I recall: Facebook's dilemmas over how/what to censor have existed since very early in the company's history. Internally, FB developed a policy document - separate from the public one we see - which would be applied by both human and AI enforcers. The employment of those human enforcers could be considered unethical in itself (at least without providing adequate therapy). Over time, the secret policy document has grown substantially to account for a lot of edge cases. While terrorist/graphic violence is normally censored, there are provisions for allowing content that would otherwise not be seen/known if not shared on FB (e.g., content that governments seek to cover up). Still, there is considerable internal disagreement among executives at FB when it comes to handling edge cases.
The fact that this secret document (and the processes that define and apply it) has so much power in shaping public perception, politics, etc. with so little transparency is perhaps the most disturbing part of this.