Naturally occurring arsenic in groundwater affects millions of people around the world. In order to assess the extent of this problem, we have used machine learning to create a global prediction map of groundwater arsenic greater than 10 mg/L. The resulting model includes already known arsenic-affected areas as well as new potentially contaminated areas. Combining the global arsenic prediction model with household groundwater-usage statistics, we estimate that 94–220 million people are potentially exposed to high arsenic concentrations in groundwater. As groundwater is increasingly utilized to support a growing population and buffer against increasing water scarcity due to a changing climate, this model will help raise awareness, identify suitable areas for safe wells and guide where testing for arsenic should be prioritized.