It just doesn't add up.
[1] https://spectrm.io/insights/blog/messaging-app-statistics-mo...
[2] https://www.androidauthority.com/imessage-big-deal-guide-308...
Be ready for an expansion of their fear-mongering ad campaigns about how anyone in your office or home can read your messages when you send them unencrypted.
I think you're missing how information is extracted and what E2EE means. Most of the machine learning models work off of metadata and so the actual content of the messages isn't exactly necessary. Yes, they can help, but it is also often a lot of noise. They've probably found and experimented with WhatsApp and found that they are able to extract enough information through the metadata that they can still maintain high accuracy on inferences. So why not make the PR stunt and move towards encryption? It actually does do some good for people (also note that Twitter has been talking about E2EE messages, though remember that they partnered with OpenWhisper very early on...).
This is really the main difference between the technical aspects of Signal and WhatsApp. Both use the same E2EE protocol but Signal does not capture metadata. We need to remember that metadata is essentially all the information a private investigator would have on you if they followed you around but could not hear the actual content of any conversations you had with the people that you meet. But they would know who you meet, how often, where, etc. Metadata is pretty useful data.
- all numbers in every user's contact list
- the membership of every group
- the sender and recipient of every message
- when every message was sent, received, and viewed or played
- the dates and times that every user has been present
- the type of every message sent (text, sticker, video, voice recording, poll, etc.)
- the size of every message sent (to the nearest 16 bytes)
- the date, time and duration of all calls made between users
And probably other metadata too. Basically everything except for the message content.
The client also uploads crash dumps (along with a copy of the log output), and periodically sends usage statistics too.
The value? They don't need to secure their databases against external AND internal attacks (believe it or not, fb has protections in place so that employees don't randomly access user data), they can more easily be compliant with the never-ending stream of regulations like GDPR, and... they can more easily do interop with WhatsApp (which doesn't have plaintext messaging).
so early it was pre-FB purchase?
Define "always". The very early releases of WhatsApp used plaintext. Encryption was added in 2013, and E2E (using the Signal protocol) was added in 2016.
https://threatpost.com/whatsapp-adds-end-to-end-encryption-t...
> “Over the past year, we’ve been progressively rolling out Signal Protocol support for all WhatsApp communication across all WhatsApp clients. This includes chats, group chats, attachments, voice notes, and voice calls across Android, iPhone, Windows Phone, Nokia S40, Nokia S60, Blackberry, and BB10,” wrote Open Whisper Systems founder, Moxie Marlinspike.
Download signal.