There should be no way for a designer to not be able to figure out why their machine produced what it did given enough debugging.
The machine's designers must not be able to explain how their original code led to this new program.
If I'm not mistaken, this has already happened when evolutionary algorithms were applied to hardware design: some slides. The author of the program has no idea how the resulting circuit worked.
The presumption is that where you valued the information enough to lock it in a safe and not write down the password, that you have a good enough memory that you anticipated being able to remember it.
OR, it's a bunch of porn that you don't want anyone to see but still won't mind losing by accident.
If your memory could be read by doctors using a harmless mind-reading machine, that would be allowed, because it would be physical evidence, not testimony that might have been compelled.
Actually, given the fallibility of human memory, I don't see how you could convert imperfect memories into anything better than a "machine-forced testimony". You just can't beat entropy by using a magical machine.
BLISS is ignorance.