Who do you blame when the machine that did the crime was set loose by no one?

I've spent years chasing people through their expense reports, and every fraud trail ends at a human choice. But now the systems are making the choices, and the trail is just logs. Genuine question for anyone who's worked with automation: when something goes wrong and no human pushed the button, is the fault with the person who designed the conditions, the person who monitored the logs, or the machine itself?