Throughout Midwestern forests, invasion by the exotic plant garlic mustard [Alliaria petiolata M. Bieb. (Cavara & Grande)] has become increasingly problematic. A multi-criteria risk model was developed to predict invasion in the Upper Peninsula of Michigan where garlic mustard is present but not yet widely distributed. The model uses geographic information system (GIS) data to predict the risk of invasion at three phases: introduction, establishment and spread. Known occurrence data for garlic mustard were used to validate the model predictions, with 89% of points correctly identified at moderate to high risk for invasion. The risk model predicted 13% of the Upper Peninsula to be at high risk and 33% at moderate risk for establishment of garlic mustard. Field sampling of randomly generated points across the Upper Peninsula provided only two additional observations of garlic mustard presence. The low encounter rate during field sampling may indicate that garlic mustard has not yet reached its full invasion potential in the Upper Peninsula. This presents an opportunity to use the model predictions and associated risk maps for monitoring and management in a relatively uninvaded region.