“Even HRW’s own posts seeking examples of online censorship were flagged as spam, the report said.”
Human Rights Watch documented more than a thousand removals, suspensions and shadow bans across more than 60 countries. It blames automated moderation and government pressure. Meta’s answer is that 1,000 examples is a rounding error, which works when the only one who can count the rest is Meta. Instagram translated “Palestinian” plus “Praise be to Allah” as “Palestinian terrorists,” and nobody at Meta caught it.