NASA publishes open AI toolset

NASA is putting more of its artificial intelligence work into public view, with open models and code aimed at planet hunting, solar forecasting, and spacecraft operations. (science.nasa.gov) In astronomy, these systems sift through huge streams of images and sensor readings to flag patterns humans would miss in time. NASA said ExoMiner++, an updated open-source package from Ames Research Center, was trained on data from both Kepler and the Transiting Exoplanet Survey Satellite and identified 7,000 exoplanet candidates in an initial run. (science.nasa.gov) A solar storm is a burst of activity from the Sun that can disrupt satellites, radio links, and power systems on Earth. NASA said its Surya model was trained on 9 years of Solar Dynamics Observatory observations and produced solar-flare forecasts two hours ahead that beat prior benchmarks by 16%. (science.nasa.gov) NASA and its partners are also building internal tools for engineers and mission teams, not just science pipelines. Element 84 said ChatGSFC, a chat system built for NASA on the open-source LibreChat platform, is used by more than 7,000 people across the agency for work including risk analysis and mission planning. (element84.com) For spacecraft operations, the basic problem is coordination: several satellites have to share tasks, avoid collisions, and still follow a human operator’s goals. A NASA TechPort description for a swarm coordination and planning effort said the software combines task allocation, path planning, collision detection, and a human-swarm interface for missions involving multiple vehicles. (techport.nasa.gov) The common thread is that NASA is favoring software other researchers can inspect instead of treating every model as a black box. NASA said ExoMiner++ can be freely downloaded from GitHub, and the agency said Surya’s model weights and code were released through Hugging Face and GitHub. (science.nasa.gov, science.nasa.gov) That approach lines up with a broader agency push to use artificial intelligence in “weather, mission planning, and more,” while stressing safe deployment in mission work. NASA’s public artificial intelligence page says the agency uses these systems to analyze data, support research, and help spacecraft operate autonomously when they are out of contact with Earth. (nasa.gov) At Goddard Space Flight Center, that policy has been formalized. A public white paper titled “NASA GSFC AI Strategy,” acquired on April 8, 2025, calls artificial intelligence a “foundational technology” for the center’s long-term goals and lists internal assistants among the expected applications. (ntrs.nasa.gov) Not every tool in the conversation is equally public today: ExoMiner++ and Surya have public repositories, while ChatGSFC is described as available only inside NASA even though it is built on open-source components. The result is a mixed model in which NASA is opening code where it can and keeping some operational systems behind agency walls. (github.com, github.com, element84.com) The release pattern is less about a single product launch than about how NASA wants artificial intelligence used in high-stakes settings: visible enough to inspect, test, and adapt before it is trusted with science and flight decisions. (science.nasa.gov, science.nasa.gov, nasa.gov)