From the US to Colombia to China, millions of public surveillance cameras are at the core of crime prevention strategies. Yet, little is known on the effects of surveillance cameras on criminal behavior. We study an installation program in Medellín and find that quasi-random allocation of cameras led to a decrease in crimes and arrests. With no increase in monitoring capacity and no chance to use camera footage in prosecution, the results suggest offenders were deterred rather than incapacitated. We find no evidence of close range negative or positive spillovers after the installation of the cameras.