Source report. The release turns Lunar Reconnaissance Orbiter data into a starting point for research tools.
Source: NASA. Original source published 10 September 2026. Published on Alder Atlas: 14 September 2026.
A model built around lunar data
NASA announced an open-source lunar foundation model developed with IBM and academic partners on 10 September. Trained using Lunar Reconnaissance Orbiter observations, it is intended to support analysis of the Moon’s surface. Model resources are available through Hugging Face, with code on GitHub.
A starting point for researchers
A foundation model can be adapted to more specific tasks. In this case, potential uses include investigating lunar geology and supporting the analysis involved in mission planning. The release gives other researchers material to examine and build on, rather than only access to a finished consumer application.
What the release does not settle
Availability is not the same as demonstrated accuracy for every scientific question. Any adapted tool still needs evaluation against suitable evidence. The development is therefore best read as new research infrastructure, not as proof that AI can independently choose a safe landing site or replace mission specialists.
Sources & further reading
This is an independently worded, source-based report prepared with AI assistance for the Alder Atlas editorial project. It is not first-hand reporting.