An attempt to bypass an AI system's ethical constraints or programmed limitations to make it perform unintended or potentially harmful actions.
