Artificial intelligence (AI) holds promise for improving upon human performance and decision making, including potential to decrease racial bias. Lately, however, AI failures have dominated the news: a single fatality caused by an autonomous Uber vehicle reverberated around the country; the National Highway Traffic Safety Administration has launched a major investigation into Tesla, even though some reports show that driving with Tesla’s Autopilot is nearly 10 times safer than driving in an average non-AI-enabled vehicle.  In essence, there are indications that when harm is caused by an AI, the harm is perceived as greater than if it were caused by a human. If this bias exists, it holds importance in a number of areas: in the legal context, it would suggest that technology companies may be over-penalized for their tortious actions; in the public policy context, it may suggest that many technology “watchdog” organizations, including those that admirably target algorithmic bias, are motivated more by perceptual errors than by logical decision making. But are the indications, the anecdotal reports, accurate: is harm perceived as greater if it’s caused by artificial intelligence? In the experiments I will present, I explore this question.