I read somewhere that WhatsApp uses some AI to detect CSAM and send content automatically and without user consent to Facebook for review. Has this been confirmed?
^ This is the Head of WhatsApp at Facebook specifically calling out Apple for their client-side scanning, pointing out that Apple should instead have a way to report content; so, while he technically could be lying, I doubt that's true.
The wired/gizmodo article tries its best to imply that it does.
the main claim comes from one of the moderators saying that: "the ai keeps sending us mundane pictures, like kids in bathtubs". This has allowed people to claim that facebook are scanning every image/message. Where as, obviously given the paltry number of moderators, your going to us AI to triage messages. Otherwise you'd have to spend more cash on staff
I think there is another bit where someone says they can trace the journey of the image. Its pretty trivial to do this with metadata, given that whatsapp leaks a fair bit of metadata its not suprising.