Disclaimer: The views expressed in this work represent the personal views and conclusions of the authors writing in their personal capacity and do not reflect the official policy or position of the Air Force, the Department of Defense, the U.S. Government, or the Stanford Space Law Society.

(To access the Print Version, please click here)

Access in, from, and to space underpins critical U.S. national interests. Specifically, this includes achieving space superiority, enabling global mission operations, and ensuring assured space access. Beyond national security, space is increasingly vital for the commercial sector, driving economic prosperity and innovation, and directly impacting the American way of life by providing unfettered access to space-based applications that fulfill consumer needs and requirements. Space-based capabilities are becoming more inherent and technologically integrated into military, commercial, and civilian applications as well as joint service processes. These capabilities empower the joint warfighter with threat reporting data and intelligence to quickly access near worldwide coverage to denied areas of interest, conduct intelligence, surveillance, and reconnaissance (ISR) operations, target identification, and monitor adversary command and control (C2) activities at various orbits.[1][2]

Artificial intelligence (AI) models have the potential to synthesize big data, enhance analytic capabilities in space-based threat reporting for analysis, and address the scalability challenges of sensor data overload. However, a gap remains between the analytic processes, sensor-data scalability, and analytic tradecraft needed to establish meaningful connections for the warfighter in a timely manner to understand and counter adversary intentions and behaviors. Closing this gap is essential to prevent operational surprises in the space domain. It underscores the need for future AI capabilities to enhance sensor fusion, predictive analysis of space events, adaptive collection/monitoring, and, most importantly, intelligence decision support.

On January 23, 2025, President Trump signed Executive Order 14179, Removing Barriers to American Leadership in Artificial Intelligence, to reinforce U.S. global leadership in AI and promote responsible innovation.[3] Effectively harnessing AI has the potential to significantly enhance mission effectiveness by integrating both existing and emerging capabilities.[4] In this context, the United States Space Force (USSF) and the Department of the Air Force (DAF) must build the requisite AI capacity to improve analytic workflow and close critical gaps to secure the U.S. competitive advantage in space.

Assessing Adversary Intentions: Challenges in the Space Domain

“We cannot, as a country or a service, miscalculate the capabilities, force posture, or intentions of our potential adversaries. We must have timely and relevant indications and warnings to help us avoid operational surprise in crisis where appropriate to take defensive actions. This means we need to have access to and invest in actionable space domain awareness”

General B. Chance Saltzman, Chief of Space Operations, At the Mitchell Institute 3rd Annual Spacepower Security Forum[5]

Accurately assessing adversary intentions in space—specifically, how they might disrupt, degrade, or deny U.S. access to critical space-based systems—represents one of the most demanding and essential tasks for analytic tradecraft and all-source intelligence. The Intelligence Community (IC) analytic standards guide analysts in the systematic and rigorous analysis of information and actionable insights that adhere to the five Analytic Standards, including the nine Analytic Tradecraft Standards.[6] Effective application of these analytic standards requires objectivity, comprehensive source utilization, political independence, and timely delivery.[7]

Analysts face a primary challenge in this increasingly complex workspace: the timely identification of relevant, diagnostic information. Access to timely information gives commanders and intelligence staff a real-time understanding of developing events. As a result, established situational awareness and understanding allows them to anticipate and deter gray zone activities against U.S. and allied interests.[8]  For instance, the Chinese Communist Party (CCP) continues to accelerate its military-civil fusion strategy, leveraging China’s commercial and military defense industrial sectors.[9][10] This strategy takes advantage of a rising volume of China’s space activities and launches, directly straining analytic tradecraft. It also challenges analysts to accurately differentiate and assess China’s commercial space operations from military activities and initiatives emerging in the space domain.[11]

Operational Environment (OE) Shift: The Strategic Need for AI and Space-Based Technologies

A significant development transforming the space OE is the advancement of counterspace capabilities by U.S. competitors—particularly China and Russia—designed to capitalize on U.S. dependence on space-based assets and services. These nations have a strategic understanding of the U.S.’s heavy reliance on space for warfighter precision technologies, national commerce, and overall economic stability through its global space-based infrastructure and networks. Furthermore, China and Russia recognize this dependence and view it as an asymmetric vulnerability that they can exploit. Additionally, they are actively pursuing coercive tactics and innovative strategies across space and cyber domains, contributing to analytic uncertainty by operating “below perceived thresholds for U.S. military action and areas of responsibility across different parts of the U.S. Government.”[12] These competitors actively seek new counter-space capabilities to disrupt, degrade, or deny U.S. space-based critical operations and services.

Key AI Approaches to Enhance Analytic Outcomes: Building Analytic Capacity in the Field of AI to Secure U.S. Competitive Advantage in Space

Maintaining constant situational awareness in the space domain presents significant challenges for analysts, operators, and analytic processes. The spatial relationships between assets, whether friendly or adversarial, are dynamic and constantly changing. This complexity is further compounded by the ambiguity of gray zone activities, the rapid pace of technological advancements in space, and the evolving OE. These factors, combined with the overwhelming volume of sensor data, make it increasingly challenging for analysts to accurately identify, observe, and, most importantly, characterize and assess potential threats in a timely manner based on existing ground and space-based sensor data networks.

In this regard, AI models can play a pivotal role by helping analysts build and enhance analytic capacity across every phase of the intelligence cycle to include multimodal sensor tasking for efficient data processing and exploitation, threat-focused analysis and production, data collection across multiple sources to validate findings and create operational threat pictures, and real-time alert notifications to support intelligence dissemination efforts.[13][14]

The emergence of Foundation Models (FMs), i.e. large deep learning neural networks, offers a “transfer learning” approach for intelligence processes through AI.[15] Transfer learning focuses on extracting “knowledge” learned from a particular task, such as space object recognition, and applying that “knowledge” to another task, such as activity recognition in videos.[16] The emergence and homogenization of FMs can help analysts and data scientists create a “real-time operational picture.” They enable the performance of multi-source intelligence and multimodal tasks, i.e. tasks that combine different types of data from different sources in real time, such as text, images, video, or audio. This allows for the generalization of new insights from the Unified Data Library (UDL)[17] repository or training data, all while reinforcing baseline analysis to better address complex problems.[18][19]

Moreover, FMs can incorporate Large Language Models (LLMs) to identify causal relationships between variables and events.[20] LLMs are a specialized subset of FMs that can be fine-tuned for specific tasks to enhance Natural Language Processing capabilities. Through aligning syntax, semantics, and context, LLMs can generate more sophisticated, contextually aware, and task-specific AI-driven outputs. For instance, LLMs can be used to improve intelligence analysis through hypotheses generation driven by historical data and events, red teaming, or high-impact/low-probability analysis.[21] Furthermore, LLMs can be combined with Bayesian networks, which model probabilistic distributions of relationships, to offer a powerful hybrid approach for addressing analytic uncertainty and conditional probabilities.[22] Bayesian networks leverage probabilistic data to reveal correlations between events, even when information is incomplete, ambiguous, and uncertain. This makes them vital for analysts navigating complex developments. LLMs further refine the Bayesian network analysis by improving interpretability and synthesizing narratives from these patterns, leading to more informed insights for analysts.

Unlike FMs that can extend the concepts of LLMs through multimodal processes across diverse datasets and formats, LLMs primarily specialize in large-scale linguistic or language processing. However, the choice between using FMs or a specialized LLM depends on the specific requirements of the application, task, or data type, as shown in Figure 1.

 

Figure 1. AI Model Feature Classification, Impacts, and Distinctions for Multimodal Processing and Intelligence Fusion in the Space Domain[23][24][25]

However, an emerging capability of LLMs is their capacity to connect and enable data collaboration and knowledge transfer between different models and systems. This networking potential was showcased at the Air Force’s BRAVO hackathon in spring 2023.[26] This event unlocked opportunities for teams to generate, create, and incorporate faster and more comprehensive insights for space operations. Furthermore, novel methods facilitated two-way communication between specialized LLMs, leading to enhanced situational awareness through networking for space operators.[27]

Conclusion

On October 24, 2024, the first-ever National Security Memorandum in AI was released, outlining a whole-of-government approach to the innovation, security, and international consensus of AI norms and ethics to advance the national security mission.[28] Additionally, the White House Office of Management and Budget, in coordination with the Assistant to the President for Science and Technology, issued two revised policies to facilitate responsible AI adoption to improve public services: M-25-21 and M-25-22.[29] These new policy memos highlights the intention of the U.S. to maintain global dominance in AI. M-25-21 helps agencies adopt AI innovation quickly and responsibly, while M‑25-22 provides clear guidance for acquiring best-in-class AI in a competitive and responsible manner.[30] In doing so, the Department of Defense must leverage AI models to improve and enhance counter-space strategies and analytic solutions that can effectively streamline massive datasets, refine analytic tradecraft, and improve all-source intelligence capabilities in support of space operations.

As global competitors such as China and Russia continue to pursue new methods to exploit vulnerabilities in U.S. space-based infrastructure and operations, the practice of all-source intelligence and analytic tradecraft in the space domain must evolve to keep pace with the scale and complexity of sensor data. Adapting to the challenges of large-scale data streams and dissecting the intricacies of adversarial intentions and behaviors will depend heavily on the integration of AI models, applications, and decision-support solutions. This marks only the beginning of a broader transformation in all-source intelligence within the space domain.

* Dr. Marquay Edmondson is a Lieutenant Colonel in the United States Space Force and National Security Affairs Fellow at the Hoover Institution, Stanford University.

[1] Clayton Swope et al, Space Threat Assessment 2024 (Center for Strategic & International Studies, April 17, 2024), https://csis-website-prod.s3.amazonaws.com/s3fs-public/2024-04/240417_Swope_Space_Threat_0.pdf?VersionId=DDeJ0EkYnF5W7POfMJHVGjkxEVeTx3o0.

[2] United States Department of Defense, 2022 National Defense Strategy of the United States of America (U.S. Department of Defense, 2022), https://media.defense.gov/2022/Oct/27/2003103845/-1/-1/1/2022-NATIONAL-DEFENSE-STRATEGY-NPR-MDR.PDF.

[3] Executive Order 14179, 90 Federal Register 8741 (January 31, 2025).

[4] Office of Management and Budget, Executive Office of the President, Memorandum for the Heads of Executive Departments and Agencies (April 3, 2025), https://www.whitehouse.gov/wp-content/uploads/2025/02/M-25-21-Accelerating-Federal-Use-of-AI-through-Innovation-Governance-and-Public-Trust.pdf.

[5] Unshin Lee Harpley, “Saltzman Pushes Need for ‘Actionable’ Space Domain Awareness,” Air & Space Forces Magazine, March 27, 2024, https://www.airandspaceforces.com/space-force-space-domain-awareness-saltzman/.

[6] Office of the Director of National Intelligence, Intelligence Community Directive 203 Technical Amendment, Analytic Standards (2015), https://www.dni.gov/files/documents/ICD/ICD-203.pdf.

[7] Id.

[8] Gray zone activities are “coercive approaches that may fall below perceived thresholds for U.S. military action and across areas of responsibility of different parts of the U.S. government.” (Source: United States Department of Defense, supra note 2 at 6.)

[9] United States Department of State, Military-Civil Fusion and the People’s Republic of China, accessed August 20, 2025, https://www.state.gov/wp-content/uploads/2020/05/What-is-MCF-One-Pager.pdf.

[10] Swope et al, supra note 1.

[11] United States Department of State, supra note 9.

[12] United States Department of Defense, supra note 2 at 6.

[13] Space Training and Readiness Command, Space Doctrine Publication 2-0: Intelligence Doctrine for Space Forces (July 2023), https://www.starcom.spaceforce.mil/Portals/2/SDP%202-0%20Intelligence%20%2819%20July%202023%29_1.pdf.

[14] Space Training and Readiness Command, Space Doctrine Publication 3-100: Space Domain Awareness (November 2023), https://www.starcom.spaceforce.mil/Portals/2/SDP%203-100%20Space%20Domain%20Awareness%20%28November%202023%29.pdf.

[15] Rishi Bommasani et al, On the Opportunities and Risks of Foundation Models (Center for Research on Foundation Models, 2021), 4, https://crfm.stanford.edu/report.html.

[16] Id.

[17] UDL is a centralized repository launched by the Air Force Research Laboratory. It is designed to combine and manage data from a variety of different satellites, both commercial and military, to allow individuals to make better‑informed decisions on space-related activities. (Source: Gage Miller, “The Unified Data Library: National Asset or Security Risk?,” Space Security, June 29, 2024, https://spacesecurity.wse.jhu.edu/2024/06/29/the-unified-data-library-national-asset-or-security-risk/.)

[18] United States Department of Homeland Security Science and Technology, Foundation Models at the Department of Homeland Security: Use Cases and Considerations (April 2023), https://www.dhs.gov/sites/default/files/2023-12/23_1222_st_foundation_models_dhs_paper.pdf.

[19] Bommasani et al, supra note 15.

[20] Moritz Willig et al, “Can Foundation Models Talk Causality,” preprint, arXiv, December 23, 2022, https://doi.org/10.48550/arXiv.2206.10591.

[21] United States Government, A Tradecraft Primer: Structured Analytic Techniques for Improving Intelligence Analysis (March 2009), https://www.cia.gov/resources/csi/static/Tradecraft-Primer-apr09.pdf.

[22] Yu Feng et al, “BIRD: A Trustworthy Bayesian Inference Framework for Large Language Models,” preprint, arXiv, April 3, 2025, 1, https://doi.org/10.48550/arXiv.2404.12494.

[23] United States Department of Homeland Security Science and Technology, supra note 18.

[24] Booz Allen, “Linking Large Language Models for Space Domain Awareness,” Defense One, January 12, 2024, https://www.defenseone.com/sponsors/2024/01/linking-large-language-models-space-domain-awareness/393302/.

[25] Lisa Costa, “Schriever Spacepower Series: Dr. Lisa Costa”, moderated by Gen Kevin P. Chilton, posted November 8, 2023, by Mitchell Institute for Aerospace Studie, YouTube, https://www.youtube.com/watch?v=EdBnGYii3nA.

[26] Booz Allen, supra note 24.

[27] Id.

[28] National Security Memorandum on Advancing the United States’ Leadership in Artificial Intelligence; Harnessing Artificial Intelligence To Fulfill National Security Objectives; and Fostering the Safety, Security, and Trustworthiness of Artificial Intelligence, 2024 Daily Compilation of Presidential Documents (October 24, 2024)

[29] The White House, Fact Sheet: Eliminating Barriers for Federal Artificial Intelligence Use and Procurement (April 7, 2025), https://www.whitehouse.gov/fact-sheets/2025/04/fact-sheet-eliminating-barriers-for-federal-artificial-intelligence-use-and-procurement/.

[30] Id.