Why you need to know this: if you have ever been certain about a colleague and unable to point at a single thing they did, this explains why, and it explains why the research is stuck in the same place you are.
Most people can name somebody at work whose conduct is unimpeachable on paper and whom everybody who has worked closely with them quietly routes around, and nothing has been raised, nothing is on file, and the people who have tried to explain it out loud have generally come off worse for trying. The popular account of that person is that they are a good actor, which is satisfying and does no work; the account this article is built on is that they are exercising restraint, and that restraint is close to the hardest thing in psychology to measure, because every instrument for it takes one reading at one moment while the thing itself is made of what somebody does not do across months.
Start with what happens when somebody watches instead of asking. Three researchers gave 317 adults in Poland an evening survey every day for thirty consecutive days. Measured the usual way, by questionnaire, Machiavellianism and psychopathy overlapped by more than seventy per cent, which is why some researchers have argued the two are not separate things at all; measured day by day, against what people said they had actually done, the overlap collapsed to about sixteen per cent. What separated them was that the Machiavellian days looked like keeping a low profile to get your way and the psychopathic days looked like walking into something dangerous, and a day of manipulation predicted a day of recklessness afterwards while the reverse never held.
Then look at the study you will be quoted at. A survey of 812 students in Taiwan found that narcissism, psychopathy and everyday sadism all predicted misusing generative AI on coursework, and that Machiavellianism did not. The popular reading of that, which is that the calculating are too careful to risk it, is not what the authors said: they said such people under-report, and that five of their own Machiavellianism questions did not hang together with the rest of the scale. Another study of 283 people found the opposite, and a review of twenty-three studies links the trait to plagiarism and cheating, but not consistently to paying somebody else to do the work.
This leads to the finding that organises everything else. Your evidence is thin not because you failed to gather it, but because what you are looking at is defined by producing less of it.
Three more results sit alongside those. Across 1,438 people, high Machiavellianism combined with political skill earned high-performance ratings from colleagues, while the same trait without political skill earned low ones, so the people who get caught are largely the ones doing it badly. Across 550 people rated by 1,127 of their colleagues, the only route by which the trait costs anything runs through reputation, and political skill closes that route. And across 8,587 records from a German panel study, higher Machiavellianism went with a greater chance of being a manager, rising steadily up the scale rather than levelling off somewhere respectable.