The relationship between trust and risk is a topic of enduring interest. Although there are substantial differences between the ideas the terms express, many researchers from different disciplines have pointed out that these two concepts become very closely related in personal exchange contexts. This raises the important practical concern over whether behaviors in the widely-used “trust game” actually measure trust, or instead reveal more about risk attitudes. It is critical to confront this question rigorously, as data from these games are increasingly used to support conclusions from a wide variety of fields including macroeconomic development, social psychology and cultural anthropology. The aim of this paper is to provide cogent evidence on the relationship between trust and risk in “trust” games. Subjects in our experiment participate either in a trust game or in its risk game counterpart. In the trust version, subjects play a standard trust game and know their counterparts are human. In the risk version, subjects know their counterparts are computers making random decisions. We compare decisions between these treatments, and also correlate behavior with subjects’ risk attitudes as measured by the Holt and Laury (2002) risk instrument. We provide evidence that trusting behavior is different than behavior under risk. In particular, (i) decisions patterns in our trust and risk games are significantly different; and (ii) risk attitudes predict decisions in the risk game, but not the trust game.