Moral licensing, equivalently called "self-licensing", is the instrumental use of a Good Act to cover up a Bad Act. This paper's thesis is that "instrumental apology" i.e., bad-faith apology, is a case of moral licensing. A decision maker may issue an apology (Good Act) after committing a Bad Act, but if the decision maker uses the apology instrumentally, he or she is using the apology to justify the Bad Act. Hence, the apology is insincere. Sincerity is the fine line between a good-faith apology or, more generally, a Good Act, on one hand, and an instrumental apology or, more generally, moral licensing, on the other. In this light, moral licensing should be separated from genuine apology that attains moral equilibrium, which is called in the literature moral "self-regulation" and "conscience accounting". According to Kantian ethics, not just the consequences of an act matter, but also the sincerity with which the act was conducted. This pits Kant against the utilitarian view, which downplays intentions and focuses on consequences. We take Kant to the lab. Participants play a modified ultimatum game, where proposers in some treatments have the option of issuing apology messages and responders have both costly and costless options for rewarding or punishing proposers. We introduce different treatments of the apology message to allow responders to form doubts about the sincerity of the apology messages. Our results support the Kantian position: responders, once they become suspicious of the sincerity of the proposers' apology, exhibit "insincerity aversion" and punish proposers.