In open systems like the Internet, traditional approaches to security based on identity do not provide a solution to the problem of establishing trust between strangers, because strangers do not share the same security domain. A new approach to establishing trust between strangers is trust negotiation, the bilateral exchange of digital credentials describing attributes of the negotiation participants. This approach relies on access control policies that govern access to protected resources by specifying credential combinations that must be submitted to obtain authorization. We describe a model for trust negotiation, focusing on the central role of policies. We delineate requirements for policy languages and runtime systems for trust negotiation, and evaluate four existing policy languages for trust management with respect to those requirements. We conclude with recommendations for extending existing policy languages or developing new policy languages to make them suitable for use in future trust negotiation systems.