Virtual worlds present tremendous advantages to cyberlearning. For example, in virtual worlds users can socialize with others, build objects and share them, customize parts of the world and hold lectures, do experiments, or share data. However, virtual worlds pose a wide range of security, privacy, and safety concerns. This may lead educators to become (or not) apprehensive of the virtual worlds in using and adapting them as learning technologies. This study examined how educators perceive risks and uncertainties in virtual worlds. We also investigated how educators' level of use of virtual worlds influences their risk perception level. Our results indicate a divergence between risk perception and reality in the virtual worlds. We use the seminal risk perception model developed by Fischhoff and his colleagues, and our revision to this model to explain these results. Finally, we discuss implications of our research for education management, and make recommendations to educators and policy makers who consider using virtual worlds as a learning technology.