I don't understand why they claim that scanning messages erodes privacy: it can be done in the recipient side totally on client side. It could be as easy as a image classifier: "hey, this photo may contain unwanted content (we may be wrong), do you want to see it? Yes, No, Report, Report & Block".<p>Unfortunately Facebook do an awful job blocking harmful content: I have reported posts where there is obvious phishing, hate speech or even murder threats and... They do nothing. But if you write "tinta negra HP" (HP black ink) they ban you immediately.