WASHINGTON — As the Pentagon races to define the future of missile defense under its ambitious Golden Dome initiative, industry leaders point to what could be the most daunting obstacle facing the program: it’s not the hardware. It’s the data.
Golden Dome aims to build a layered, integrated missile defense shield spanning land, sea, air, and space — connecting existing and future systems into a real-time, AI-enabled network to counter everything from drones to hypersonic glide vehicles. The initiative, driven by a Presidential directive, has a 60-day window to produce an “objective architecture,” according to Gen. Michael Guetlein, the program’s top manager.
That architecture is far from being settled, Guetlein said at a recent industry event. “Everyone who’s telling you they think they know, do not know, including me,” he said.
Executives from L3Harris Technologies, Booz Allen Hamilton and Arcfield during a SpaceNews panel July 31 stressed the need for clear scope and requirements definition so the program can leverage existing technologies but also allow industry to inject innovations in areas like AI and machine learning that will be essential to the success of Golden Dome.
Patrick Biltgen, vice president of space AI at Booz Allen Hamilton, said Golden Dome requirements are still unclear.
“What is the point of Golden Dome? Is it to protect the homeland or is it to provide tactical capabilities or does it also do hypersonics over the ocean?” he asked. The answer to that question changes everything — scope, cost, timelines, and tech requirements. A continental shield is one thing. A global tactical umbrella? Entirely different.
Dan Knight, vice president of sensors and data integration at Arcfield, suggested that while Golden Dome is intended as a large defense for North America, it would also need to support “tactically forward units” and adapt to overseas conflicts involving drones or short-range ballistic missiles. He emphasized the potential for integrating allied missile defense capabilities, noting that harnessing the sensors and shooters of allies is desirable for a comprehensive system, despite the program being named “Golden Dome for America.”
Rob Mitrevski, President of Golden Dome Strategy and Integration at L3 Harris Technologies, highlighted that the command and control (C2) aspect is “certainly the most complex part” of the architecture.
“We’ve moved from sort of the modeling and sim discussion to the decision support discussion in real operations,” said Mitrevski. “If you look at some of the missile raids that we’ve seen in actual conflict over the last number of months and few years, you’ll know that the decision support will be of utmost importance, and AI is a very strong application of that, particularly in trying to prioritize which targets with which interceptors at which sequence of events need to happen in a very short time frame.”
The challenge becomes even more acute with hypersonic weapons. “When you’re talking hypersonics, it becomes increasingly important because of the lower latency,” Mitrevski added. “Drones are slow and intercontinental ballistic missiles are fast. But hypersonics, in their trajectory, are very difficult targets, and so AI support will be very important.”
The AI revolution in defense
Artificial intelligence is emerging as a critical enabling technology for Golden Dome, particularly in addressing data challenges and the escalating “speed of the game” in modern warfare, Biltgen said. The system’s complexity, comprising existing and new components, makes it “very hard to prescribe data and integration standards and demand that everyone integrate with them,” he said. “So that’s an opportunity for AI enabled integration.”
Advanced AI models are now capable of data translation and even writing their own services to interoperate — capabilities that were unimaginable just a few years ago. This includes automated code generation and automated testing of system integration.
However, the speed requirements are pushing the boundaries of traditional human-in-the-loop operations. Previously, public discourse always emphasized human oversight, but now, “the speed of the game is driving the requirement”for greater automation, said Biltgen. This necessitates verification and validation to trust a “full auto mode under certain circumstances,” as “the human is the weak point in the loop” when speed is paramount.
Security and workforce challenges
The AI integration comes with significant security considerations. Biltgen underscored that the government may need to train its own unique, secret models not known to adversaries, rather than relying on general-purpose open-source models that could be exploited.
“One of the areas that’s complex and potentially very expensive is the government may need to train their own unique secret models that are not known to the adversary,” Biltgen said. “Golden Dome is all about ambiguity. Are the interceptors and sensors in the right place where they can actually intercept? And if you give the adversary enough ambiguity about their success, then they can’t attack you.”
Biltgen also noted there is a significant gap in the AI workforce, which necessitates immediate collaboration with universities to create a workforce for the future.
Industry awaits clarity
The defense industry will get its next major update on August 7 at the “Golden Dome Industry Summit” conference in Huntsville, Alabama.
“I’m hoping that General Guetlein and his team come in and say, here are some things we know, here are some things we don’t know, but we’re leaving the design space open for some creative solutions,” Biltgen said.
Knight said data management will be a central issue in Golden Dome. “We have the information or we have the data that we need. It’s just not in the right places. And so going forward, I think it’s going to be essential for us to architect what is currently in place through model based systems engineering, and then build out on that.”
Mitrevski emphasized the need for government prioritization: “That’s really critical, because it is a very complex system, and the things that are ready and easy should be done, because we do need to focus on the things that are difficult to see … and connecting all these disparate pieces is really going to be the hard part.”
Overcoming AI skepticism
Despite the technical promise of AI, “some people are really weirded out by the idea that these interceptors might be AI enabled, like you’re going to have these angry space robots that are just going to chase things and intercept them,” Biltgen said. However, he argued that mindsets may need to change because with humans in the loop, it might be difficult to react quickly enough to intercept hypersonic missiles within the required timeline.