NASA is moving to treat data and software as strategic assets under a new agency directive aimed at accelerating artificial intelligence (AI) development and improving access to information across its scientific, engineering, and business systems.
Kevin Murphy, acting NASA chief data and super intelligence officer and chief science data officer in NASA’s Science Mission Directorate, said the directive from the agency’s deputy administrator is driving efforts to connect previously siloed systems and modernize the infrastructure needed to support AI.
Speaking at a GovCIO event on scaling AI and data across federal agencies, Murphy said NASA plans to expand its use of application programming interfaces to make internal information more accessible to authorized users. The effort will also extend to NASA’s publicly available scientific data.
“You can’t have super intelligence or artificial intelligence without … a really good data ecosystem,” Murphy said.
The modernization push supports NASA’s broader mission goals, including returning astronauts to the moon by 2028, advancing scientific instruments, and developing the next generation of experimental aircraft.
Murphy said achieving those goals requires better connections between wind tunnels, computing environments, engineering databases, and other systems while maintaining security.
Data governance remains a major challenge, he said. NASA, Murphy said, must manage controlled unclassified information, export-controlled data, and proprietary information as it expands data sharing and considers how sensitive information can be incorporated into AI models.
The agency is also exploring ways to use its approximately 200 petabytes of publicly available scientific data for AI development, he said. According to Murphy, NASA is working with commercial partners, including NVIDIA and IBM, on foundation models using scientific observations.
Murphy pointed to lunar imagery as one example of the technical challenges involved. Unlike Earth-based observations, lunar imaging lacks GPS positioning and requires additional calculations to account for orbital conditions, camera positioning, and lighting.
NASA is also experimenting with generative AI for physics-based design and considering AI agents for project management and configuration control.
However, Murphy said the agency has not yet resolved how to govern AI agents that can access data or perform tasks across interconnected systems.
“I think that’s still a to be solved question,” Murphy said, citing ongoing work to determine the policies, infrastructure, and management controls needed to support agent-based systems.
NASA is evaluating vendor solutions and examining the costs and operational requirements associated with deploying agents. Murphy warned that insufficient oversight could lead to unintended consequences, particularly as the agency balances security requirements with scientists’ and engineers’ ability to experiment.