It would be difficult to create such persuasive adversarial examples in this case. Humans can recognize images well, so it's obvious for us that those adversarial examples are nowhere near what neural net thinks they are. It would not be as obvious with a list of phone calls.