Personally I think we need to start utilising the safety features built into AI, to ensure that who we're talking to is a human. We'll start to have to only reply to people who talk in nsfw cursewords (like cocks), or profess their love of capybaras