As NGA directors need AI to help manage rapidly growing space-based sensor data, the National Geospatial-Intelligence Agency focuses on developing responsible but agile AI standards.
NGA Director Vice General Frank Whitworth spoke about the opportunities and challenges of NGAs and the wider Geoin community at the Intelligence and National Security Summit in National Harbor, Maryland on Thursday. Whitworth has been NGA director since 2022. Earlier this month, President Donald Trump appointed Lt. Gen. Michelle Bredenkamp to serve as the agency's next director.
Whitworth said the main issues with NGAs are the increased availability of space-based sensors and the corresponding data flooding that those sensors bring to NGAs.
“We won't go into detail about exactly what the rate of growth in terabytes from space is, but it's quite a bit over the next decade,” he said. “We don't want all that data, all those images, all those files – and they're very dense files – we don't want them to fall on the cutting room floor.”
While the National Reconnasance Office has long managed highly classified satellite constellations that provide images and other data to NGAs, Space Force deploys “proliferated architecture” of low Earth orbit satellites.
Meanwhile, NGAs are increasingly consuming images and other data directly provided by the rapidly growing commercial space industry.
NGA has long used computer vision and machine learning to manage the decades of growth in GEOINT data. But as yet another data surges onto the horizon, Whitworth declared 2025 “AI Year” as the NGA's “AI Year”. He then established new positions focusing on this issue, including Director of AI Mission, Director of AI Programs, Director of AI Standards.
An AI Standards official said, “I ensure that all the development that's going on there will be alienation from the standard that we as Americans are true.
Last year, the NGA established an initiative called Geoint AI Model Certification or Agaim. This initiative provides a standard evaluation and risk management process for AI models that can be applied to Geoint.
“We need to make sure there's no hobby stores happening here,” Whitworth said.
“We just want to avoid entering Cocom or the services or internal institutions using training data that moves away from principles, like untested, unrated training data, unrated processes, and laws of armed conflict,” he added.
At the same time, Whitworth said he wants to avoid establishing a lengthy certification process that will slow down the deployment of responsible AI models.
“We don't want to move in the direction of [authority-to-operate] “The process of everyone sitting and waiting,” he said. “We want this process to allow for self-governance, so we train Cocom and our services, trainers, train our certified people, and then we can also certify them in the right way.”
AI standard work is particularly important as NGAs and other agencies are considering adopting industry models to manage, analyze and share data.
A few years ago, the NGA adopted the Pentagon's “Project Maven” program. Maven uses machine learning to process data from multiple sources and identify potential military targets.
Whitworth said the NGA is surpassing the “targeting cycle” with its Maven program.
“We want to ensure that, like navigation safety, we bring goodness to the rest of what we're doing, like warnings,” he said.
The NGA has established a new Program Executive Office for Advanced Analytics to reduce all mission areas of the NGA.
“It's starting to bear some fruit,” Whitworth said.
Copyright©2025 Federal News Network. Unauthorized reproduction is prohibited. This website is not intended for users within the European Economic Area.
