I don't understand "capability". Surely you realize the capability is already there. Scanning an image and sending a "bad detected" message is absolutely trivial. The image classification algorithms that run locally, used for categorical search (like "dogs") are already there.
When I say capability I mean the functionality currently exists to do that, it’s implemented and functioning, and the evidence for it existing is not a surprise.
If WhatsApp has the function for Facebook to read messages, then it’s not going to be a surprise when reversing the program to see that code; but if they claim that there is no capability to do that and you discover it then it’s much more damning.
Also; if functionality exists it’s easier to argue for a scope change… much easier than arguing for new functionality to exist.
It’s the motte and bailey.