I want the real thing.
I'm not in agreement with this being ok, but if it truly is on device, it still technically can be E2EE
What do you think this software does when it finds a matching hash entry? Toss a notification to ask you nicely to pop round your nearest FBI office?
What meaning does 'end-to-end' have anymore if this applies? If Apple wrote software on-device to forward a copy of all messages prior to encryption to iCloud for 'backup', would it still be end-to-end? What if they sent it to an AdTech firm to index for interesting terms that match products you should be pitched? The software in this case is still Apple's, running on-device.
If you used iOS/iMessage, you have always trusted Apple. To be truly 'end to end' you would assume that there's no opportunity for another party to intercept your messages: the only trust would be in the keys you exchanged with your peer.
1. They are a trusted broker for iMessage key exchanges. You didn't do the key exchange, you assumed that when Apple did so on your behalf that you're really communicating with the peer you think you are.
2. They designed the iOS features that you trust keep iMessage data inaccessible to other untrusted software on your device.
3. They designed the secure enclave and make public statements that they won't compromise it for law enforcement. You trust that their deeds in private match their public statements.
I'd suggest you read up a bit on this - https://www.apple.com/child-safety/
What we're discussing is this feature:
"The Messages app will add new tools to warn children and their parents when receiving or sending sexually explicit photos.
When receiving this type of content, the photo will be blurred and the child will be warned, presented with helpful resources, and reassured it is okay if they do not want to view this photo. As an additional precaution, the child can also be told that, to make sure they are safe, their parents will get a message if they do view it. Similar protections are available if a child attempts to send sexually explicit photos. The child will be warned before the photo is sent, and the parents can receive a message if the child chooses to send it.
Messages uses on-device machine learning to analyze image attachments and determine if a photo is sexually explicit. The feature is designed so that Apple does not get access to the messages."
There are no visual derivatives, no neural hashes, and nothing is sent to apple.