Facebook decides that "their users are better served" by scanning messages locally on the client device and requesting ads based on that content, thereby introducing a privacy and information leak and potentially opening the door to other problems.
I don't necessarily think that's likely, but it's also not entirely implausible that they would do something, intentionally or not, that subverts the threat model/privacy assumptions of its users.
Facebook's interests are not aligned with users when it comes to privacy.
Does Facebook always detail their tracking and analytics platform updates? In any case, my point is that it doesn't need to be malicious for FB to invalidate the privacy expectations that users of encrypted messaging might have.
If it was malicious, they already control the client code and therefore have access to the plaintext anyway.
I don't necessarily think that's likely, but it's also not entirely implausible that they would do something, intentionally or not, that subverts the threat model/privacy assumptions of its users.
Facebook's interests are not aligned with users when it comes to privacy.