Governments around the world conduct ethics trainings with public employees to enhance public integrity. Yet, causal evidence on the effectiveness of ethics trainings to enhance integrity in government remains rare. We address this gap through a field experiment with 1,400 police officers in Bangladesh. In collaboration with one Bangladeshi police district, we randomly assign half of the district’s police officers to a state-of-the-art ethics training. Our training trains and primes each participating police officer to be an ethical leader in the police district. Through a panel of survey data and incentivized honesty game data, we assess the effect of the ethics training on shifts in the integrity-related attitudes, perceived norms, beliefs of perceived behavioral control (cf. moral efficacy), and behaviors of police officers. We find positive effects of the ethics training on perceived behavioral control and on curbing unethical behavior of police officers. Our findings have important implications for our understanding of the effectiveness of ethics trainings as a tool to enhance integrity in government, and the malleability of integrity-related attitudes, norms, control beliefs and behaviors of public employees.