Should platforms be held liable for the harms suered by users? A two- sided platform enables interactions between rms and users. There are two types of rm: harmful and safe. Harmful rms impose larger costs on the users. If rms have deep pockets then platform liability is unnecessary. Holding the rms liable for user harms deters the harmful rms from joining the platform. If rms are judgment proof then platform liability plays an instrumental role in reducing social costs. With platform liability, the platform has an incentive to (1) raise the interaction price to deter harmful rms and (2) invest resources to detect and remove harmful rms from the platform. The residual liability assigned to the platform may be partial instead of full. The optimal level of platform liability depends on whether users are involuntary bystanders or voluntary consumers, and the intensity of platform competition.