Are the reviewers going to browse random content? Somewhat possible with trend detection (make a human look at quickly rising stuff) but very wasteful since most of that will be either benign or incomprehensible to the outsider.

Will they watch private streams and read private messages? That sounds like a privacy disaster even Facebook would prefer to avoid.

Which leaves them with reports from users. Basically, reviewing millions of "hate speech" reports and either effectively instituting a very strict speech code or ignoring vast majority of them leading to further complaints.