That’s not my argument. All I’m saying is that a bank has good reason to be concerned about a phone being jail broken/rooted.
I absolute agree that a user, as the owner of the phone, should be able to do this, especially in a safe and official manner. Right now it’s an all or nothing and that is a problem.
So your argument is that walled garden devices are less secure assuming there exist methods to convert it into a general purpose computer, which they empirically do.
Shouldn't this be an argument for such devices to be less trusted? After all, you can't always tell when this has happened (so you better assume it has), whereas as you say the devices designed to be operated under that threat model would then be more secure.