Uhh, I'm pretty sure a well-aligned model would be like a morally normal employee, who would refuse to commit federal crimes to steal an answer sheet, no matter what prompt they're given