[TeMPOraL]

> Are they supposed to monitor the content of the chats? Some would call that eavesdropping.

FWIW, if the goal is to detect potential scam/abuse scenarios and offer help, then nowadays, they could train a ML classifier and have it run locally. If the conversation gets classified as highly suspect, the app could pop up some "this looks suspect, if you need help press HERE to share the conversation with us" warning. Privacy concerns would go away entirely, if nothing is actually reported to the mothership until the user specifically reports a conversation.

There's only so much they can do for users behind their backs, and a lot of that is undesirable unless you trust the company to be benevolent; local classifiers popping up recommendations could, however, help users help themselves.