Military Leaders Say AI Needs Better Data to Reach the Battlefield

Military officials say the biggest obstacle to putting artificial intelligence into combat operations isn’t developing the technology, it’s getting AI access to the data it needs while keeping it inside the right security environments.

“The toughest challenge for us is adoption,” said Robert Shaffer, deputy director of intelligence at U.S. Space Command. “It’s figuring out how to properly use it, how to do it in the right enclave and give it the access to the material that it needs in order to help support our decision-making and drive that decision advantage.” 

That issue was a recurring theme during a panel moderated by Marc Zuccola, deputy CIO at the Defense Intelligence Agency, at DoDIIS Worldwide in Tampa, Florida. Recent operations, including Epic Fury, have underscored the need to rapidly integrate data from multiple sources and deliver it to commanders and operators. Jananne Swoopes, deputy director for intelligence for the Joint Staff, said the military can’t afford to treat competition and conflict as separate data environments. 

“We need it integrated. We need it automated. We need it faster,” Swoopes said. “We need the J3, the [J4], the [J5], the [J6] to be able to talk together and share data, and we need to be able to do that very, very rapidly. All the commanders have talked about the advantage. You know, we have data advantage, and we have decision advantage, and we have workflow advantage. But we need that not just horizontally. We need a vertical.” 

AI is already beginning to contribute to parts of the process, she said, including bringing data to analysts faster and supporting target systems analysis. The longer-term goal is to use AI throughout the targeting cycle, from intelligence collection through assessing whether a kinetic or non-kinetic effect achieved its intended result.  

The technology could also help sift through the growing volume of data generated by new sensors. Shaffer said AI could help identify “the signal and the noise,” while also supporting collection and processing, exploitation and dissemination. 

But deploying AI at scale will require new approaches to cybersecurity. Elizabeth Durham-Ruiz, director of C4 and CIO at U.S. Strategic Command, said the command is working to establish a registry of AI models and determine who has access to them. 

“There are so many well-intentioned individuals who’ve been helping us, so we’ve got to have a registry,” Durham-Ruiz said. “We also have to have some methodology of tracking access to that, whether it’s some kind of credential access management capability.” 

The officials ultimately pointed to a broader cultural shift as necessary for AI adoption. And resilience must also be built into the architecture before a crisis occurs.  

“You can’t grow your PACE plan in the middle of a crisis,” Durham-Ruiz said, referring to primary, alternate, contingency and emergency communications plans.