Lunar AI

NASA and IBM are collaborating on an open artificial intelligence foundation model designed for lunar science. Trained on 17 years of observations from NASA missions, the model is intended to interpret and map the Moon’s surface at a level of detail that can support both research and future exploration. Its applications include identifying likely water ice, cataloging craters, tracing volcanic features and recognizing other geological formations across large areas of the lunar landscape.

The model builds on the idea of adapting foundation-model techniques, widely used for language and imagery, to scientific data. Instead of serving as a general chatbot, it is meant to help researchers process extensive remote-sensing records, generate maps and detect patterns that could otherwise require substantial manual analysis. Making the system openly available could allow universities, commercial space companies and international research teams to test it, improve it and apply it to new lunar datasets. The effort also reflects growing interest in AI as a tool for planning missions and evaluating potential landing or resource-extraction sites, especially as NASA and other space agencies prepare for more sustained activity around the Moon.