Almost a year later, on a hot day in the high summer of 2025, I walked into NGA headquarters at Fort Belvoir Army Base in northern Virginia. It was my second visit to the intelligence agency HQ, and I wanted to know why Whitworth had changed his mind, how widespread Maven had become, and how Maven’s new subordinates saw the risks and rewards of introducing AI into the military stream.
By then, Whitworth had become such a fan of AI that his organization was providing machine-generated intelligence reports to US judges that “no human hands” had touched. And the NGA had launched a $708 million contract for data labeling to support Maven computer vision models, the largest appeal in US history, which would eventually go not to Billionaire Alexander Wang’s Scale AI but to Empowered Intelligence, a startup aimed at hiring people on the autism spectrum expert in good job referrals.
My visit required the rigors of any meeting in an intelligence agency. background checks and reviews; no cell phones, laptops, or smart watches allowed; and one more curious step: writing down not only the model and model, but also the serial number written on my tape recorder, which I decided not to use again for any interview after the visit.
This building was the temple of geospatial intelligence, or GEOINT, the movement of detailed analysis associated with locations on maps. A reflective glass mesh covered with nearly 2,000 concrete triangles covered the blast-resistant surface, as if each one was trying to turn a different area. More than 8,500 employees worked at headquarters, but I was there to meet four specific NGA officials. Each, in their own way, was heavily involved in the development, standards, and spread of Maven. I was told, it was unprecedented for them all to gather in one room to tell a reporter about Maven, and I was curious to hear what was at stake for them.
“This is our reputation on the line,” Whitworth told me in an interview. After seeing how easy it was to integrate the system into combat scenarios, it didn’t take long for him to change his mind: “I started to really believe in it.” Rather than being timid about ushering in a new era of AI warfare, its midwives wanted their names stamped on it. Others had been “sophisticated” in seeking credit, an NGA official said. I wondered if the NGA wanted its fair share, remembering that some advising the second Trump administration wanted to take control of Maven and AI away from the NGA and back to the Pentagon. “Nobody can claim credit for this. It’s huge.”
NGA officials walked me through Maven’s development since the agency took over two years ago. Five of Maven’s eight projects, including analyzing drone feeds and satellite images, ended up with NGA. Whitworth wanted to expand the scope and capabilities of his agency based on the expansion of ubiquitous global sensors. AI relied on data, and that required global tracking to deliver it. While the NSA can listen to the world, the NGA can watch it. Whitworth made it clear he wanted to do it in minute, constant detail—exploring the whole world, all the time. NGA previously gave me a demo showing how AI can signal military construction in China—like the arrival of a new railway station at a missile base. NGA tracked all movements at 49,000 airports around the world. Whitworth even wanted to put a GPS, or similar navigation system, on the moon. And if the GPS jammed or was hacked, he wanted other ways to plan space, too: NGA was developing digital maps to draw on magnetism, gravity, remote sensing, aerial navigation, and elevation. “From the bottom of the sea to the sky,” went the new mantra he launched in 2023. America’s warhorse wanted to be omniscient, omnipresent, and omnipotent.
