We elicit individual preferences over social risk. We identify the extent to which these preferences are correlated with preferences over individual risk and the well-being of others. We examine these preferences in the context of laboratory experiments over small, anonymous groups, although the methodological issues extend to larger groups that form endogenously (e.g., families, committees, communities). Preferences over social risk can be closely approximated by individual risk attitudes when subjects have no information about the risk preferences of other group members. We find no evidence that subjects systematically reveal different risk attitudes in a social setting with no prior knowledge about the risk preferences of others compared to when they solely bear the consequences of the decision. However, we also find that subjects are significantly more risk averse when they know the risk preferences of other group members.