On December 9, 2025, United States War Secretary Pete Hegseth announced the launch of GenAI.mil, a platform placing the power of artificial intelligence (AI) into the hands of every U.S. warfighter. The platform brings one of the most capable frontier models to every government workstation across the department. Eventually, the platform will add three — and perhaps more — foundational models to its site, arming military professionals with tools to enhance productivity and multiply output.
Space professionals may wonder whether a similar renaissance will transform space systems — several of which are decades old and in desperate need of a technical refresh. In small pockets across the U.S. Space Force and U.S. Space Command (USSPACECOM), this work is being done.
AI for SDA? Why?
Space Domain Awareness (SDA) could be the greatest beneficiary of AI technology. SDA underpins almost every other function in space — from missile warning to environmental monitoring, to space control and orbital warfare. It is a highly complex process. Practitioners of this mission may describe it as determining the who, what, when, where and why a space object may behave the way it does using tiny specks of information gathered from across the widest domain around the globe.
However, if one were to examine this SDA kill chain, we can see that AI can augment discrete segments to yield exponential results. For example, the necessary process of performing conjunction analysis can take hours or days, depending on the number of space objects involved, the orbital regime the conjunction occurs in and the number of sensors that can see where the conjunction would occur.
How does one prioritize what conjunctions are likely? A trained AI agent can perform these duties, sorting through thousands of potential cases to call the operator’s attention to the scenarios that are likely to occur. The operator then can focus on these cases, running additional analyses as needed, and notify the owner or operators of those satellites to prevent a collision.
This process is not as easy as it seems. Training data is often sparse, and when available, it is frequently classified, complicating the appropriate training of these conjunction or detection models. While there is utility in using synthetic data to better train models or AI tools to help extrapolate existing data to theorize what may have happened in a sensor gap, there is no true replacement for hard data. Commanders performing space defense missions — more often than not — will prefer to accept some data over no data.
Therefore, the exponential growth and interest in commercial companies wanting to contribute to this mission presents unique opportunities to provide higher-fidelity data, which would make space a safer domain to operate in. Providers such as Slingshot Aerospace, LeoLabs and BlackSky bring new capabilities to the table, adding different sensing modalities and providing more coverage in space than ever. The Space Force and other consumers of space data would benefit greatly in not dismissing the power of commercial data.
Moreover, the overlap between government and commercial entities could be highly beneficial for space enthusiasts. In particular, access to commercial data facilitates data sharing with allies and partners because there are fewer classification and foreign disclosure hurdles. Some of the most impressive work is being done by the Joint Commercial Office, a team that augments the U.S. National Space Defense Center’s mission to protect and defend U.S. and allied space capabilities exclusively through the use of commercial data.
AI for the Space Domain
Here are five ways artificial intelligence can help improve space domain awareness, as outlined in an issue brief by U.S. Space Force Maj. Christopher Huynh that was published in June 2025 by Georgetown University’s Center for Security and Emerging Technology.
Catalog maintenance
Accelerate the association of unknown space objects to known orbit tracks for lists such as the Space Force’s space-track.org.
Orbit determination
Reduce the time it takes to predict a satellite’s future position to seconds instead of minutes or hours.
Conjunction assessment
Narrow thousands of object pairs to the few that are most likely to collide, then rapidly assign each
a risk score.
Data integration
Harmonize disparate types of data from a variety of sensors.
Sensor scheduling
Help set priorities for overtasked satellites and other sensors.
In parallel, the Space Force could benefit from a technical refresh of its government SDA architecture. Some may describe it as archaic, nonmodular, network-centric and stovepiped. Day by day, this system is being burdened by the exponentially growing number of tracked objects in space. New systems built on modernized software architectures, such as the Advanced Tracking and Launch Analysis System, can alleviate some of these issues.
Sophisticated cybersecurity is a critical component of successful space operations. In the Space Force’s efforts to modernize its SDA architecture, there must be a mindfulness toward implementing robust cybersecurity tools such as zero-trust practices; modernized identity, credential and access management protocols; and decentralized data-centric security.
AI agents aboard the satellite?
One of the best use cases for AI in space is putting the technology onboard the satellites. Some of the most promising areas of AI development for space lie in the miniaturization of hardware, combined with more robust data, and the steady maturation of object-based sensing algorithms. Though space is a unique domain, the process of building robust models suitable for satellite survivability holds the same principles as training a robust earthbound system.
These technologies help push computation to the edge — to the satellite in space — enabling it to perform all the necessary calculations needed for autonomous responses to adversaries on orbit. Until quite recently, these complex calculations relied on first sensing the threat, computing possible response options using powerful computers on the ground, and then uploading a series of commands for the satellite to execute.
These critical minutes or hours could determine the life or death of a satellite. If a satellite possessed a combination of the three pieces of the AI triad on board — algorithms, computing power and data — it could significantly improve its chances of survival against a novel or unforeseen threat. This could prove highly valuable in situations where a satellite may be out of contact with no possible intervention by a human operator.
In a study by the China Aerospace Studies Institute at Maxwell Air Force Base in Alabama, researchers concluded that the Chinese Communist Party’s (CCP) “lack of consistent access to a worldwide network of ground sensors” pushed it instead to design an SDA architecture that is space-based, rather than ground-based, from the beginning. In contrast, the U.S. did not take such an approach. The early days of our SDA architecture have roots in missile warning, using ground radars to detect intercontinental ballistic missiles from the former Soviet Union. As space lift technology improved and the first satellites were launched into space, it evolved to track space objects, painting the foundation for SDA.
Moving forward, such architecture may not be enough. The U.S. and allied nations would benefit from bolstering its on-orbit SDA sensor coverage. This would provide resilience and better coverage of difficult-to-track regions of space, such as geostationary orbit and cislunar space.
Homing in on satellite technology, this coverage would benefit from novel detection techniques such as pose estimation. Pose estimation, traditionally used to track the arm and leg movements of humans, is quickly adapted to track the orientation, shape and movement of satellites in space. This subset of computer vision AI technology is proving to be a game-changing approach to tracking objects in space using space-based sensors. Teams such as the Space Rendezvous Laboratory at Stanford University in California provide some of the most advanced research in this critical field.
Economies of scale will matter. As pose estimation technology matures, acquisition professionals must be mindful of cost. Procuring expensive technologies in space — in hardware or software — may not be the best answer. We must be ready to account for attrition of satellites should conflicts extend into space.
Why stop there? The possibilities of AI applications onboard satellites are immense. Some essential capabilities that researchers and space practitioners also must consider are technologies that make satellites more efficient and resilient. Models or agents that are capable of autonomously detecting and responding to satellite anomalies, such as link failures, are necessary. AI-enabled satellites could be developed that are capable of formation flying in space — similar to technology being developed for the U.S. Air Force’s Close Collaborative Aircraft combat system or the U.S. Marine Corps’ experimentation with small, unmanned aerial systems.
What happens to the human?
The roles may change, but humans still will play a central role in space operations. AI will make the operator’s role easier by reducing repetitive and mind-numbing tasks. It also will enhance their ability to detect microscopic perturbations in data telemetry that could indicate a threat.
In other ways, AI will make the operator’s role more difficult. With the onboarding of new capabilities, operators may be forced to study a wider variety of satellite failure modes. When a unique failure mode occurs, will the operator understand the root cause? Will the operator have the knowledge and tools to confidently and correctly respond to the failure? Operations crews will need to be ready to respond to these possibilities — at all hours of the day and night — in the absence of expert assistance.
To facilitate this challenge, there must be emphasis on operator trust, and this will need to be an enduring requirement. One method is to increase the audits and transparency of AI tools applied to space systems. Simple interfaces, such as callouts that explain how an AI agent reached a decision — and what factors it applied to its analysis — could significantly facilitate operator trust and willingness to use the tool in a time-sensitive situation. This methodology, coined Local Interpretable Model-agnostic Explanations or LIME by AI researchers, improved user trust in AI models that operators have not interacted with before.
As AI capabilities grow, military training pipelines must evolve to account for the shift. This is where the Air Force shines. Its vision of recruiting, training, retention and education created a compelling model to attract AI talent. Programs such as direct commissioning and constructive service credit present excellent pathways for aspiring Airmen or Guardians with unique skill sets — such as AI — to join technical roles. For those who are less inclined to serve in uniform, opportunities such as Gig Eagle enable civilians to take on part-time roles that fulfill Department of War needs in critical areas.
Growing a cadre of AI-minded space professionals will be a difficult but necessary task. As space becomes more contested, degraded and operationally limited, leveraging AI technologies will be a strategic imperative. It will be critical to leverage all capabilities to respond to rapidly growing and adapting adversaries.
About the author: U.S. Space Force Maj. Christopher Huynh is a space operations officer at Headquarters, United States Space Force, under the deputy chief of space operations for cyber and data. He was previously a National Defense Fellow at Georgetown University’s Center for Security and Emerging Technology and is a graduate of the Air Force Institute of Technology and
U.S. Air Force Academy.