The GSM network is the largest network on Earth, providing vital communications service to billions of people. Yet hundreds of millions of people live outside coverage of existing cellular providers. Recently, researchers have demonstrated a new model of cellular connectivity, community cellular, that has the potential to bring coverage to extremely rural populations. Although the total capital costs for these networks (under US$10,000) are affordable for rural communities, the high financial and political cost of access to spectrum makes running a legal community cellular network nearly impossible. This is despite the fact that very rural areas typically have substantial amounts of licensed spectrum that is not actively used. We define this unused spectrum as GSM whitespace. Like TV whitespaces, we argue that GSM whitespaces should be regulated for dynamic spectrum sharing, and that doing so will support the growth of community cellular networks and thereby improve rural access to communications services. We propose a hybrid sensing and database-driven spectrum sharing scheme called Nomadic GSM that provides safe coexistence between primary and secondary users without requiring coordination or cooperation from existing license holders. Nomadic GSM also mitigates concerns about "spectrum squatting" by secondaries and provides regulators visibility into and control of the spectrum usage of secondary operators. Lastly, we implement and evaluate a prototype Nomadic GSM system that is fully compatible with existing, unmodified GSM handsets and has been deployed in a real-world community cellular network in Papua, Indonesia.