If it's a consequence of a 'buggy ML classifier', well, it's FB's policy to use one for censorship.
You can't launder accountability with an 'It's AI' black box.