Deep Sensing, the New Military AI? Why Deep Sensing Should be on the Radar

[Klaudia Klonowska is a PhD Researcher at the Asser Institute and the University of Amsterdam, and is serving as Managing Director of the Manual on the International Law Applicable to Artificial Intelligence in Warfare.

Sofie van der Maarel is Assistant Professor in Military Ethics and Leadership at the Netherlands Defense Academy and affiliated to Radboud University Nijmegen.]

Introduction

The concept of “deep sensing” has been steadily gaining traction in military discussions and media coverage, especially in the U.S., indicating an emerging buzzword amongst marketing strategies surrounding emerging military technologies. As researchers specializing in military technology, we are particularly intrigued by this concept and its purported novelty. In this post, we unpack what deep sensing is and put it on the radar of current socio-technical legal and research agendas. 

Our argument is that deep sensing represents the new trend in military AI. By analyzing the addition of “deep” to sensing, we draw parallels with the well-known narratives surrounding AI-enabled warfare – namely, the drive for greater precision, speed, and scalability. Proponents of deep sensing projects offer similar promises, including the creation of “transparent battlefields”. Despite these underlying similarities, advancements in deep sensing capabilities may have flown under the radar for those following AI developments in the military domain. By strategically labeling new projects as “deep sensing”, there seems to be an invocation of neutrality and objectivity – but this should not be overlooked by those with a keen interest in military technology, as it may obscure the more complex, underlying issues at play.

The Role of AI in Sensing Technologies

Sensing technologies are not a novel development in military settings. For years, military intelligence has relied on these tools – such as satellites and drones – to enhance situational awareness. However, sensing technologies are increasingly intertwined with AI-enabled data processing and analysis systems. As the military gathers more data, it becomes ever more dependent on AI to process and filter “actionable” information. Vice-versa, the greater the integration of AI into military operations, the stronger the need for high-quality, up-to-date data. While this dynamic is well-understood, what is important is the framing of new platforms and systems as purely “sensing” platforms. This framing downplays their potentially critical reliance on AI technologies, and obscures the ethical and legal implications they present. For example, Lockheed Martin’s F-35 is presented as an AI-enabled fighter jet, described as a “fusion” of sensors designed to provide “a single integrated picture of the battlefield that greatly enhances awareness, survivability, and lethality.” This form of marketing that places emphasis on “fusion of sensors” may cause one to overlook the importance of this data collection effort to train and experiment with AI-enabled technologies, as well as the inherent dependence not only on sensors but also AI to produce the “picture of the battlefield”. 

Key Projects in Deep Sensing: HADES and TITAN

The first operational imperative for the Army of 2030 is to see and sense farther, and more persistently at every level across the battlefield than our enemies. So how are we going to do that? We will need to collect and analyze unprecedented amounts of raw data from many different types of sources.

These remarks, delivered by former U.S. Army Secretary Christine Wormuth at the Annual Defense Programs Conference in March 2023, highlighted the Army’s broader ongoing modernization efforts, particularly in intelligence, surveillance, and reconnaissance (ISR). A central focus of these initiatives is the development of “deep sensing” capabilities – Artificial Intelligence (AI)-driven data processing technologies designed to enhance the Army’s ability to collect and analyze battlefield data with unprecedented precision.

Two key projects lie at the heart of this initiative: the High Accuracy Detection and Exploitation System (HADES) and the Tactical Intelligence Targeting Access Node (TITAN) ground station. As part of the HADES program, the Army equips aircraft for ISR missions, delivering “deep sensing” capabilities to “the warfighter”. But the programs also offer highly advanced data processing capabilities to inform target acquisition and nomination. Meanwhile, TITAN, marketed as the Army’s “next-generation” ISR ground station, is designed to deliver “deep sensing to field artillery”. Collaboration with private sector companies, such as Palantir and Anduril Industries, further underscores the reliance on machine learning to process sensor data and “bring actionable targeting information to the tactical edge”. Palantir, a major defense contractor, plays a pivotal role by providing AI-enabled services, including those delivered to the Ukrainian armed forces. 

We bring attention to deep sensing developments to make sure that legal scholars and practitioners do not overlook them, recognizing their role in already ongoing discussions on sensing and AI in the military. Discussions within the security domain are often highly politicized, as seen in the use of Lavender systems in Gaza. We therefore argue that framing these technologies as natural “sensors” represents a seemingly strategic shift that diverts attention away from critical discussions on algorithms driving warfare. Below, we further unpack the concept of deep sensing and explore its broader implications.

What is “Deep” in Deep Sensing?

The origins of deep sensing can be traced back to the U.S. Army’s priority “to provide the ability to see farther, see more, and see more persistently than our enemies”. In October 2023, the U.S. Army offered a vague definition of the term in Army Field Manual (FM) 2-0: Intelligence, where deep sensing is described as “the employment of capabilities beyond the division coordinated fire line to collect data and information that supports targeting, situational understanding, or decision making.” In a 2024 statement, an Army representative elaborated further, stating that deep sensing is about “being able to see over the horizon, (…) see on the other side of the mountain or seeing deep into enemy territory.” However, these definitions fail to accurately capture the built-in reliance on AI applications required to filter, process, and analyze the increased volume of sensor data associated with deep sensing capabilities.

Take, for instance, the example of HADES, the U.S. Army Futures Command’s program that operationalizes the promise of deep sensing into action. In November 2024, HADES acquired its first Bombardier Global 6500 aircraft. The aircraft embodies the Army’s aspiration to see more, higher, faster, and farther. It is one of many platforms that, together, will deliver aerial, space-based, and satellite data. The aircraft’s ability to operate at high altitudes is a key feature that makes it suitable for deep sensing, meaning that it can create visibility “beyond the division coordinated fire line”. This alludes to the Army’s concept of “deep area,” referring to regions over the horizon. 

Another key aspect that makes HADES capable of deep sensing is its ability to “fill in the gaps” of information. HADES ensures that sensor data collection continues even in degraded communications environments – such as when communication is jammed or otherwise interrupted by enemy forces. The “deep” in deep sensing, therefore, also seems to refer to the system’s resilience, enabling continuous surveillance despite adversarial interference. 

In addition to expanding its range of vision, the Army is using deep sensing projects to extend its “vision” over the horizon. Recent interviews with HADES project members indicate that another key objective is to provide “indicators and warnings”, develop “patterns of life for target development”, and offer capabilities to “detect, locate, identify, and track critical targets”. This emphasis on prediction underscores again that discussions on deep sensing are intricately linked to broader debates on military AI.

In a similar vein, the TITAN project promises to process sensor data from “space, high altitude, aerial and terrestrial layers”. By integrating data from various sensors into a single platform – known as data fusion – it leverages machine learning to transform this data into actionable intelligence that enhances situational awareness and informs decision-making. The integration of sensors goes beyond ground or air operations, supporting multi-domain operations, which have become a high priority for armies engaged in contemporary warfare. 

As these projects illustrate, having sensors alone is often insufficient. The deployment of such technologies generates overwhelming amounts of data, making algorithmic systems essential to translate that data into actionable intelligence and predict events in spatial and temporal gaps. This becomes particularly necessary, for example, when environmental conditions compromise the quality of images. 

This is how deep sensing evolves into new trends in discussions on military AI. Systems like HADES and TITAN are presented as sensing at higher altitudes, faster speeds, and over greater distances, generating vast amounts of data that could never be analyzed by humans in real-time. To transform these large amounts of data into informed decisions, situational awareness, or actionable intelligence – the primary goal of military sensing – data collection alone would not be enough. Additional capabilities are then sought to convert these large data sets into meaningful insights for informed decision-making. 

The collaboration between the HADES and TITAN projects and private companies further showcase the reliance on AI-enabled capabilities. These deep sensing initiatives are made operational through AI techniques developed by Palantir and its partners. For example, the TITAN project works with a Palantir-led consortium of private companies – including Anduril Industries, Northrop Grumman and L3 Harris – “to produce and deliver decision-quality targeting information to warfighters”. 

Deep Dive into Deep Sensing: What’s at Stake?

What’s at stake when we are talking about deep sensing? There are three main implications to consider regarding deep sensing projects and their entanglement with AI applications. 

The first implication is that deep sensing, as exemplified by the HADES and TITAN projects, increases the military’s reliance on AI systems. These systems are essential for analyzing the enormous amounts of data collected by integrated sensors. This reliance inherently necessitates placing the analysis of systems marketed as “deep sensing” platforms within broader legal debates on military AI. Ongoing legal debates raise concerns over the speed of AI-driven decision-making, which could undermine decision-makers’ ability to perform critical precautions during attacks. Similarly, the opacity of AI systems raises concerns over accountability mechanisms and the ability of commanders to ensure that deployed systems comply with existing obligations under international humanitarian law. 

A second implication is that as more data is gathered, the military’s dependency on companies that integrate and analyze this data for decision-making will only increase. This is closely tied to critical discussions on AI and the intersection of the military-industrial-venture capital complex. The platforms and software provided by these companies are necessary for processing and fusing of the vast amounts of data gathered through deep sensing – data that is, by design, more expansive, faster, and farther-reaching. Recent studies on the involvement of AI defense contractors in military affairs have moreover shown the emergence of prototype warfare ideas, a trial-and-error mindset, and a greater tolerance for risk. It is crucial that deep sensing technologies and platforms, such as HADES and TITAN, are scrutinized whenever delivered in cooperation with private actors, to ensure accountability and oversight. 

A third implication is that deep sensing systems should be understood within the broader context of sensing as a non-neutral and subjective endeavor. Research in media and critical security studies has emphasized the political implications of choices about who is sensed, when, and how. These decisions lead to considerations of visibility and invisibility in security contexts, including concerns related to how individuals who are sensed are portrayed. For example, HADES is fundamentally a project that enables aerial surveillance, which brings to the forefront common concerns about remote security practices, privacy of individuals, and related ethical and legal questions of operators perceiving those they sense from above.

In these ways, monitoring deep sensing projects requires addressing their inherent connection to ongoing legal debates on military AI, the military-industrial-venture capital complex, and the non-neutrality of sensor technologies.

Looking Ahead 

This piece suggests that lawmakers and researchers, as well as others following developments in military AI, should maintain a broad focus and remain vigilant to newly emerging buzzwords that may sideline or strategically avoid critical debates. Our objective was to put deep sensing projects on the agenda, stimulating further exploration of what exactly these developments entail. Terms such as deep sensing are often enveloped in a haze of complex abbreviations, technical language, military jargon, and secrecy. It is crucial to move beyond marketing strategies and unpack what exactly these technologies are and interrogate their political and legal implications. Our piece serves as an entry point for this discussion – one that will only become more pertinent in the years ahead. 

Photo attribution: “Responsible AI in the Military Domain – REAIM 2023” by Ministerie van Buitenlandse Zaken is licensed under CC BY-SA 2.0