The Ethical Demarcation Between Anticipatory Warfare and Pre-Crime: Operationalizing International Humanitarian Law through Geospatial Artificial Intelligence

The Ethical Demarcation Between Anticipatory Warfare and Pre-Crime: Operationalizing International Humanitarian Law through Geospatial Artificial Intelligence

Introduction: The Epistemological Divide in Algorithmic Warfare

The rapid integration of artificial intelligence (AI) into military operations and state security apparatuses has precipitated a profound epistemological and ethical crisis in the interpretation of International Humanitarian Law (IHL). As states increasingly rely on predictive algorithms to navigate the complexities of modern conflict and national security, a critical demarcation has emerged between two fundamentally incompatible paradigms: individualized predictive policing (often termed "pre-crime") and theater-level anticipatory forecasting. The former, exemplified by domestic law enforcement initiatives and authoritarian surveillance architectures, attempts to probabilistically map human intent and behavior. These systems frequently collapse under the weight of mathematical fallacies, generating catastrophic human rights violations and civilian harm. The latter paradigm, operationalized through conceptual frameworks analogous to Geospatial Artificial Intelligence for Theater Exploitation (GAITE), bypasses the subjective cognitive and behavioral domains entirely. Instead, it relies on sensor-derived physical realities—such as the tracking of troop movements, the construction of logistical nodes, and systemic spatial alterations—to predict macro-level events. This report exhaustively examines the mathematical, technical, and legal boundary separating these two paradigms. By mapping GAITE’s reliance on objective spatial and spectral data—derived from intelligence architectures like the Space-based Machine Automated Recognition Technique (SMART) and the Constellation Observing System for Meteorology, Ionosphere, and Climate (COSMIC)—against the rigid IHL mandates of distinction, proportionality, and precaution, this analysis demonstrates how anticipatory warfare can be ethically and legally operationalized. The core thesis posits that by shifting the object of algorithmic analysis from human subjectivity to physical hardware and systemic spatial transitions, theater-level forecasting circumvents the base-rate fallacy that plagues individualized targeting. Furthermore, the report critically evaluates the claim that accurately modeling the second- and third-order systemic impacts (reverberating effects) of a strike makes military engagement fundamentally more ethical. It finds that by grounding the doctrines of targeting and anticipatory self-defense in objective physics rather than subjective threat perception, advanced geospatial AI aligns algorithmic warfare with the rigorous demands of the laws of armed conflict.

The Mathematical and Ethical Collapse of Pre-Crime

The catastrophic failure of individualized predictive policing and algorithmic targeting systems is rooted in a fundamental mathematical flaw that renders them incapable of aligning with the legal frameworks of IHL or international human rights law. From the Los Angeles Police Department's defunct LASER program to China's Integrated Joint Operations Platform (IJOP) used in Xinjiang, and extending to contemporary military targeting architectures like "The Gospel" and "Lavender," these systems attempt to divine future actions from behavioral metadata1.

The Base-Rate Fallacy in Behavioral Prediction

When AI is deployed to predict rare, individualized events—such as a specific person's intent to commit a terrorist act, lead a subversion, or participate directly in hostilities—the statistical scarcity of the event inherently skews the algorithm's predictive validity4. The mathematical underpinning of this failure is governed by Bayes’ Theorem, which calculates the conditional probability of an event5. The theorem is expressed as: ![][image1] In the context of predictive targeting, ![][image2] represents the base rate: the actual, ground-truth probability that any given individual in a population is a lawful military target or an imminent threat. ![][image3] represents the probability of the algorithm flagging an individual (the test positive rate), and ![][image4] represents the algorithm's sensitivity (true positive rate). Because the base rate of legitimate targets ![][image2] in a dense civilian population is exceedingly low, even an algorithm boasting a 99% accuracy rate will produce a volume of false positives that massively eclipses the true positives4. This is not a technological limitation that can be refined with more data; it is an irreducible mathematical reality governing mass screenings for rare problems4. Consequently, systems relying on behavioral profiling force operators into an unacceptable ethical choice: either accept a massive number of false positives (resulting in unlawful civilian casualties or mass incarcerations) or generate excessive false negatives4. The application of this mathematics in domestic spheres, such as LASER or IJOP, results in the systemic harassment and detention of marginalized populations. When translated to the military domain, it results in the unlawful application of lethal force.

The Erosion of Distinction and Algorithmic Harm

When systems utilize "pattern-of-life" metadata—such as communication habits, movement routines, social affiliations, or even the sharing of physical infrastructure—to infer human intent, they fundamentally violate the IHL principle of distinction1. IHL requires an absolute distinction between combatants and civilians; in cases of doubt, a person must be presumed to be a civilian7. Pre-crime systems criminalize the statistical probability of behavior, effectively nullifying the presumption of innocence and stripping civilians of their protected status based on algorithmic guilt by association7. The deployment of behavioral predictive systems introduces pervasive forms of harm characterized in contemporary scholarship as "algorithmic beforemath" and "aftermath"10. Beforemath refers to the psychological, societal, and behavioral degradation caused by living under ubiquitous surveillance, where civilians alter fundamental survival behaviors to avoid triggering a lethal algorithm. They may abandon critical civilian infrastructure or cease communal interactions out of fear that their pattern of life will intersect with a statistical threat model10. Aftermath harm includes the compounding, long-term trauma of unpredictable, algorithmic violence that shatters the social contract10. By relying on subjective behavioral inputs to generate targets at scale, these systems automate the violation of IHL's duty of constant care. Rather than viewing proportionality as a subjective, context-sensitive legal judgment, this approach seeks to render human lives computationally manageable, reinforcing the flawed belief that ethical and legal dilemmas can be resolved through quantified casualty thresholds11.

GAITE and the Shift to Spatio-Temporal Determinism

To circumvent the ethical and mathematical voids of predictive policing, theater-level forecasting systems like GAITE necessitate a profound epistemological shift: moving the analytical gaze from human "software" (attitudes, opinions, ideologies, intent) to physical "hardware" (machines, geological transformations, cosmic processes, and structural engineering events)13.

The IARPA SMART Architecture

The technical foundation of this physical forecasting is exemplified by the Intelligence Advanced Research Projects Activity (IARPA) Space-based Machine Automated Recognition Technique (SMART) program14. SMART aims to automate the global-scale detection, classification, and monitoring of large-scale anthropogenic activities—specifically heavy construction, land clearing, and logistical deployment—using multi-source, multi-temporal satellite imagery14. Unlike predictive policing, which searches for abstract human intent, SMART focuses exclusively on spatial and temporal reasoning over long time scales to track concrete physical changes on the Earth's surface18. SMART operationalizes this through methodologies known as Broad Area Search (BAS) and Activity Characterization (AC)14. BAS entails processing a sequence of multi-spectral imagery—sourced from constellations like Landsat, Sentinel-2, and WorldView—to identify the spatio-temporal extents of activities of interest across vast geographic regions18. The system relies on "Site Models," which are data formats (often structured in GeoJSON) that spatially and temporally define the boundaries of a large-scale physical change14. In the SMART framework, a site of interest must be larger than 8,000 square meters, ensuring that the algorithm is tracking systemic, macro-level infrastructure rather than individualized human movements14. Once a physical site is localized via BAS, Activity Characterization classifies the progression of the physical event into deterministic, temporally bounded phase labels14. For heavy construction and military staging, these phases include:

By utilizing advanced frameworks like the Wide-Area Terrestrial Change Hypercube (WATCH), forecasting systems can ingest geo-referenced, multi-spectral data and adapt to varying ground sampling distances (GSD), missing sensor readings, and temporal gaps19. This establishes a physics-based, deterministic mathematical model. The construction of a forward operating base, the stockpiling of mechanized armor, or the development of a subterranean command bunker are bound by the rigid, unbreakable laws of physics, supply chain logistics, and engineering timelines. The base rate for these events in a contested theater is not an infinitesimally small probability subject to Bayes' trap; it is an observable, trackable physical reality that unfolds over time.

Multi-Domain Synthesis via COSMIC and Geospatial Sensoring

This objective modeling is exponentially enhanced by fusing SMART's terrestrial data with diverse spatial and spectral datasets, such as those provided by the COSMIC (Constellation Observing System for Meteorology, Ionosphere, and Climate) program and commercial aerospace entities like Spire Global, GeoOptics, and OroraTech21. COSMIC and its successor, COSMIC-2A, utilize radio occultation to provide highly dense, precise data on atmospheric conditions, space weather, thermospheric density, and ionospheric interference21. When combined with SMART's geospatial intelligence (GEOINT), systems can mathematically forecast the operational capacity of an adversary across all domains. For instance, AI-driven space weather forecasting can predict solar flares, cosmic radiation, and geomagnetic storms that degrade GPS and high-frequency communications21. If an adversary's physical logistics (tracked by SMART) are combined with environmental degradation metrics (tracked by COSMIC), a theater-level AI can forecast exactly when and where an enemy force will possess the physical capability to launch an offensive, or conversely, when their communications will be uniquely vulnerable to electronic warfare. The algorithm does not predict if the adversary wants to attack; it calculates the precise timeline of when the adversary's physical capacity to execute the attack will reach maturation17.

Operationalizing IHL: Distinction, Precaution, and Material Scope

The distinction between probabilistic human targeting and deterministic physics modeling has profound implications for the operationalization of International Humanitarian Law. By limiting its analytical scope to objective spatial realities, GAITE perfectly aligns with the material scope of the laws of armed conflict, fundamentally rehabilitating the principles of distinction and precaution that pre-crime algorithms destroy.

Upholding the Principle of Distinction

Article 51 of the 1977 Additional Protocol I (AP I) to the Geneva Conventions prohibits indiscriminate attacks, demanding that parties direct their operations exclusively against military objectives8. A military objective is legally defined as an object which by its nature, location, purpose, or use makes an effective contribution to military action, and whose total or partial destruction offers a definite military advantage. By leveraging SMART's phase-based site models, GAITE algorithms trace the physical lifecycle of an object from undisturbed earth to completed infrastructure14. Rather than inferring the purpose of a facility based on the behavior, metadata, or mobile-device tracking of the people inside it—a methodology fraught with lethal error—the system verifies its military nature through its engineering. It analyzes the specific spatial dimensions of a runway, the thermal signatures of specific industrial manufacturing, the thickness of concrete pours indicating blast resistance, or the defensive berms of a staging area14. This shifts the burden of proof from probabilistic human association to empirical structural realities, ensuring that civilian objects remain legally protected.

The Duty of Constant Care and Precaution

IHL demands that military commanders take all feasible precautions to verify that targets are military objectives and to minimize civilian harm11. The use of AI in targeting has been heavily criticized for eroding this duty, as commanders often succumb to automation bias, deferring to the machine's recommendation without independent verification11. However, a multi-modal hypercube analysis (such as WATCH) provides a verifiable, objective audit trail of the target's physical transformation over time20. When a commander is presented with a GAITE recommendation, they are not viewing an abstract "threat score" assigned to a human being; they are reviewing a time-series visual and spectral history of physical construction18. This allows for meaningful human control12. The commander can independently verify the spatial bounds, observe the transition from "Site Preparation" to "Active Construction," and legally confirm the military nature of the objective before authorizing force12. In this capacity, the AI serves as a true decision-support system that enhances, rather than replaces, the human legal judgment required by IHL11.

Proportionality and the Calculation of Reverberating Effects

The most complex mathematical, ethical, and legal threshold in modern warfare is the rule of proportionality, codified in Article 51(5)(b) of AP I. An attack is strictly prohibited if it "may be expected to cause incidental loss of civilian life, injury to civilians, damage to civilian objects, or a combination thereof, which would be excessive in relation to the concrete and direct military advantage anticipated"25.

Direct vs. Indirect (Reverberating) Collateral Damage

Historically, military commanders assessed proportionality based almost entirely on the immediate, direct effects of a kinetic strike—measuring the physical blast radius, thermal effects, and primary fragmentation zone25. Under this traditional view, the incidental damages considered were only those occurring in the immediate vicinity of the impact25. However, modern legal scholarship, supported by the International Committee of the Red Cross (ICRC), unequivocally requires commanders to account for the indirect or "reverberating" effects of an attack25. Reverberating collateral damage refers to the knock-on effects occurring across a causal chain containing multiple steps. For instance, if Missile A destroys Power Station B, which subsequently cuts electricity to Water Purification Facility C, leading to an outbreak of waterborne disease D in a civilian population downstream, the ultimate civilian harm constitutes reverberating effects25. For a commander to be held legally liable for excessive collateral damage under IHL, the reverberating effects must be "reasonably foreseeable" at the time the attack was planned25. The more causal steps that exist between the initial impact and the final civilian harm, the harder it is to establish foreseeability. Yet, the increasing experience of armed forces in urban combat, and the known interconnectedness of essential services, makes it objectively foreseeable that damage to essential infrastructure will have severe reverberating impacts on public health25. This is precisely where human cognitive limitations fail and where GAITE's mathematical modeling becomes a legal imperative. Human commanders cannot accurately compute the cascading failures of modern, interconnected urban infrastructure in real-time. GAITE, drawing upon vast troves of spatial, spectral, and infrastructural network data, maps these complex causal chains, transforming unforeseeable variables into quantifiable, expected outcomes.

Internal vs. External Proportionality in Dual-Use Infrastructure

The ethical demarcation of GAITE is most visible in its handling of dual-use infrastructure—objects that simultaneously serve civilian and military purposes, such as an electrical grid, a bridge, or a cyber-telecommunications hub. When evaluating a strike on a dual-use target, advanced IHL interpretations require two distinct levels of proportionality assessment: External and Internal30. External Proportionality addresses the physical harm caused to civilians and civilian objects outside of the targeted dual-use objective itself30. For example, if a cyber operation or kinetic strike physically destroys a dual-use server farm, external proportionality measures the loss of life among civilians standing in the adjacent buildings due to a kinetic blast or secondary explosion30. It requires one or more steps of physical causation affecting the surrounding physical environment. Internal Proportionality represents a highly nuanced legal doctrine. It measures the balance between the military advantage gained and the expected civilian harm caused specifically by the "destruction of the civilian part of the object" or the "ending \[of\] its civilian use or function"30. This calculation requires zero steps of external physical causation because the harm stems directly from the impairment of the object itself. Internal proportionality is further categorized:

1. Narrow Internal Proportionality: This applies when civilian activities take place physically inside the attacked object, and its impairment directly stops those activities (e.g., striking a dual-use government building, physically preventing civilian administrative functions inside)30.

2. Broad Internal Proportionality: This applies when civilian activities occur outside the targeted object, but rely entirely on the targeted object's functioning output. For example, striking a dual-use power grid cuts off the "civilian fraction" of electricity. The subsequent failure of a nearby hospital relies on this internal fraction of lost power30.

Under frameworks like the Tallinn Manual 2.0, the potential "deprivation of functionality" caused by an attack on dual-use infrastructure (whether kinetic or cyber) must be treated as damage to civilian objects and explicitly included in the proportionality assessment30. A pre-crime algorithm, obsessed with identifying human threats, is blind to infrastructural dependencies. GAITE, however, models the city or the theater as a living thermodynamic and digital system. By utilizing COSMIC environmental data, SMART geospatial mapping, and network theory, the AI mathematically calculates the precise fraction of internal proportionality14.

The Critical Evaluation: Does Modeling Reverberating Effects Make Strikes Ethical?

A core methodological question arises: Does accurately modeling the second- and third-order systemic impacts of a strike actually make military engagement fundamentally more ethical and compliant with IHL? Critics might argue that forecasting systemic harm simply gives militaries a more accurate body count, risking the normalization of massive collateral damage if a commander deems the military advantage high enough. However, a critical evaluation of IHL reveals that the integration of systems like GAITE fundamentally constrains the use of force. Under the standard of a "reasonable military commander," a long causal chain of infrastructure failure (e.g., a power grid failure leading to a sanitation failure leading to disease) might previously have been dismissed as legally unforeseeable, insulating the commander from liability for the resulting civilian deaths25. Because GAITE accurately models these cascading failures, the reverberating effects legally transition into the realm of expected incidental harm. The commander is stripped of the defense of ignorance. They are now legally obligated under Article 51(5)(b) to factor these projected deaths into the proportionality calculation25. If the aggregate harm—comprising direct blast deaths, internal proportionality (the loss of civilian infrastructure functionality), and reverberating systemic deaths—is excessive compared to the military advantage, the strike must be aborted25. In this sense, by mathematically verifying the interconnected fragility of urban warfare, GAITE paradoxically serves as a massive constraint on military force. It forces the military apparatus to confront the true, holistic cost of a kinetic engagement. Accurately modeling these impacts does not simply make warfare more efficient; it embeds the humanitarian imperatives of IHL directly into the algorithmic architecture, enforcing restraint and demanding the selection of less destructive means9.

The Material Scope of IHL (Ratione Materiae)

It is vital to note that IHL's ratione materiae threshold for incidental damage strictly requires death, physical injury, or the physical destruction of objects25. The mere deprivation of a right or service—such as cutting off civilian access to the internet or television broadcasts, as analyzed in the context of the 1999 NATO bombing of the RTS broadcasting station in Belgrade—does not constitute a violation of proportionality if it does not directly lead to physical harm25. GAITE's physics-based modeling is optimally suited for this legal nuance; it does not model abstract societal grievances, but strictly calculates the physical degradation of life-sustaining infrastructure. This ensures that the legal threshold of physical harm remains the sole, rigorous metric for canceling a strike, preventing the dilution of IHL into subjective political calculations.

Redefining Imminence: The Doctrine of Anticipatory Self-Defense

Beyond the conduct of hostilities (jus in bello), the epistemological shift embodied by GAITE fundamentally restructures the legal framework for the initiation of force (jus ad bellum), specifically regarding the highly contentious doctrine of anticipatory self-defense.

The Caroline Doctrine and the UN Charter

Article 2(4) of the United Nations Charter strictly prohibits the threat or use of force in international relations, essentially eliminating the concept of a "just war" based on political disputes31. The sole exception for unilateral action is found in Article 51, which preserves the "inherent right of individual or collective self-defence if an armed attack occurs"31. The temporal constraint of Article 51—specifically, whether a state must wait to absorb the devastating blow of an actual armed attack before responding—remains one of the most fiercely debated issues in international law31. Customary international law, rooted in the 1837 Caroline incident, allows for anticipatory self-defense if the threat of an armed attack is "imminent." The Caroline doctrine establishes that the necessity of self-defense must be "instant, overwhelming, leaving no choice of means, and no moment for deliberation," and the response must be strictly proportional to the threat31. It is universally accepted that preemptive strikes—the use of force against distant, speculative, or potential threats that have not yet materialized into an imminent attack—remain strictly illegal under international law33. For example, the 2003 US invasion of Iraq, often justified under an expanded notion of preemptive self-defense against potential weapons of mass destruction, heavily strained the international legal consensus because it lacked the requisite threshold of an ongoing or truly imminent threat35.

Resolving the Imminence Paradox through Spatial Determinism

The ethical and legal demarcation between illegal preemption (pre-crime at the state level) and legal anticipatory self-defense hinges entirely on how one defines and empirically proves "imminence"31. Authoritarian security models and pre-crime algorithms attempt to prove imminence by profiling the psychological intent or ideological trajectory of an adversary's leadership. This is inherently subjective, open to gross political manipulation, and legally insufficient to trigger the Caroline exception35. GAITE resolves this legal impasse by shifting the proof of imminence from subjective intent to objective spatial determinism14. Using SMART phase-classification models14, an AI system monitors the physical assembly of an adversary's offensive capabilities over time. For instance, if satellite multi-spectral imagery identifies the "Site Preparation" phase of mobile ballistic missile launchpads, the threat is nascent, but a strike is not yet justified14. When the system detects the "Active Construction" phase, the threat is developing. However, when the AI's temporal modeling proves that the physical logistics, fueling infrastructure, and command-and-control nodes have reached full integration and deployment—when the laws of physics dictate that the weapon system is operational and maneuvering into a launch posture—the threat becomes mathematically and physically imminent14.

The "Armed-Attack-Initiation" Precedent

Japan's formulated "armed-attack-initiation" doctrine serves as a vital legal precedent for this application of AI38. Under Article 51, Japan may exercise self-defense only when an armed attack occurs. In 1970, the Japanese government determined that a state does not need to suffer actual kinetic harm for an armed attack to have "occurred" in a legal sense; the initiation of an armed attack is sufficient. Initiation is defined not by mere likelihood or preparation, but by a "high probability" based on substantial, objective grounds regarding the means and patterns of attack38. GAITE mathematically quantifies this "high probability of initiation." By providing an undeniable, sensor-verified audit trail of the adversary’s physical state transitions—corroborated by SMART spatial bounds and COSMIC radio occultation data—GAITE allows a defending state to present empirical proof that an armed attack has been physically initiated31. The integration of artificial intelligence in this manner does not constitute a novel type of State behavior requiring a rewrite of the jus ad bellum33. Rather, it provides the evidentiary mechanism to satisfy existing law. Anticipatory self-defense is thus rescued from the realm of political paranoia and ideological pre-crime36; it becomes a verifiable algorithmic calculation of physical inevitability.

Synthesis and Conclusion

The integration of artificial intelligence into military targeting and forecasting demands a rigorous adherence to both mathematical validity and international law. The ethical demarcation between individualized pre-crime and theater-level anticipatory forecasting is absolute, defined by the fundamental object of analysis and the mathematical frameworks employed. Predictive policing and algorithmic targeting systems that attempt to infer human intent from pattern-of-life metadata—such as LASER, IJOP, and experimental military systems—are structurally flawed. They collapse under the statistical realities of the base-rate fallacy, mathematically ensuring an overwhelming ratio of false positives. Legally, they inherently violate the IHL principle of distinction, relying on probabilistic guilt to strip civilians of their protected status, and inflicting profound societal harm through algorithmic beforemath. Conversely, systems architected upon the principles of the GAITE concept—utilizing robust spatial and spectral architectures like SMART and COSMIC—operate strictly within the deterministic realm of physical reality. By mapping the undeniable physical transitions of hardware, logistics, and heavy construction through Broad Area Search and Activity Characterization, these systems shift the paradigm from predicting subjective intent to tracking objective capability. When mapped against the mandates of IHL, theater-level physical forecasting uniquely operationalizes the most challenging aspects of the law of armed conflict. By mathematically calculating the internal proportionality of dual-use objects and projecting the reverberating, systemic impacts of a strike, systems like GAITE do not grant militaries a license for broader destruction. Instead, by rendering complex causal chains of civilian harm mathematically foreseeable, the AI strips away the defense of ignorance. It legally obligates commanders to factor the true holistic cost of a strike into their proportionality assessments, thereby enforcing restraint and fulfilling the duty of constant care. Furthermore, GAITE's reliance on spatio-temporal determinism rescues the doctrine of anticipatory self-defense from political manipulation. By providing an objective, sensor-verified audit trail of an adversary's physical state transitions, it grounds the Caroline threshold of "imminence" in physical inevitability, satisfying frameworks like the armed-attack-initiation doctrine. In an era where the character of war is increasingly shaped by algorithmic speed, the alignment of forecasting with physics-based, macro-systemic realities remains the sole viable pathway to ensure that the application of force remains legally justified, highly precise, and ethically constrained.

Works cited

1. Comprehensive Toolkit for Defence Ethics \- DCAF – Geneva Centre, https://www.dcaf.ch/sites/default/files/publications/documents/DCAF\_DefenceEthicsToolkit\_EN20260517.pdf

2. Understanding artificial intelligence ethics and safety \- The Alan Turing Institute, https://www.turing.ac.uk/sites/default/files/2019-06/understanding\_artificial\_intelligence\_ethics\_and\_safety.pdf

3. Using Behavioral Indicators to Help Detect Potential Violent Acts \- RAND Corporation, https://www.rand.org/content/dam/rand/pubs/research\_reports/RR200/RR215/RAND\_RR215.pdf

4. Three Corrections \- by Vera Wilde \- Wilde Truth \- Substack, https://wildetruth.substack.com/p/three-corrections

5. KORFF \- FREE Paper on Core Issues in the PNR Case \- Data protection and digital competition, https://www.ianbrown.tech/wp-content/uploads/2021/12/KORFF-FREE-Paper-on-Core-Issues-in-the-PNR-Case.pdf

6. Dissecting Discrimination \- University of St.Gallen, https://ux-tauri.unisg.ch/EDIS/Dis5029\_embargoed.pdf

7. APPLYING EXISTING LEGAL FRAMEWORKS TO PREDICTIVE POLICING WITH ARTIFICIAL INTELLIGENCE \- ORBilu, https://orbilu.uni.lu/bitstream/10993/60344/1/Blount\_thesis\_final.pdf

8. Understanding Risks and Mitigating Civilian Harm in Urban Operations \- Canada.ca, https://www.canada.ca/en/army/services/canadian-army-journal/articles/2024/21-1-muhammedally-mitigating-civilian-harm.html

9. Understanding Risks and Mitigating Civilian Harm in Urban Operations \- Canada.ca, https://www.canada.ca/content/dam/army-armee/documents/caj-jac/en/caj-21-1/CAJ\_21.1\_ACCESSIBLE\_Understanding\_Risks\_Migitating\_Civilian\_Harm\_In\_Urban\_Ops\_EN.pdf

10. Full article: Tracing algorithmic harm: from innovation to deployment and impact on civilians, https://www.tandfonline.com/doi/full/10.1080/21624887.2026.2648442

11. The erosion of human(e) judgement in targeting? Quantification logics, AI-enabled decision support systems and proportionality assessments in IHL | International Review of the Red Cross | Cambridge Core, https://www.cambridge.org/core/journals/international-review-of-the-red-cross/article/erosion-of-humane-judgement-in-targeting-quantification-logics-aienabled-decision-support-systems-and-proportionality-assessments-in-ihl/31024C3473211FB84FD535BC815ADC03

12. Killing by Algorithm: The Crisis of Algorithmic Warfare on International Humanitarian Law, https://www.researchgate.net/publication/405023662\_Killing\_by\_Algorithm\_The\_Crisis\_of\_Algorithmic\_Warfare\_on\_International\_Humanitarian\_Law

13. From Social Entities To Ubiquitous Systems \- BIOS, https://bios.gr/new-version/uploads/files/creative-explorations-Publication.pdf

14. Repository associated with the IARPA SMART Heavy Construction dataset. \- GitHub, https://github.com/pubgeo/IARPA-SMART

15. SMART \- IARPA, https://www.iarpa.gov/research-programs/smart

16. Automated global-scale detection and characterization of anthropogenic activity using multi-source satellite-based remote sensing imagery \- SPIE Digital Library, https://www.spiedigitallibrary.org/conference-proceedings-of-spie/12525/1252502/Automated-global-scale-detection-and-characterization-of-anthropogenic-activity-using/10.1117/12.2663071.short

17. Geospatial-Temporal Sensemaking of Remote Sensing Activity Detections with Multimodal Large Language Model \- arXiv, https://arxiv.org/pdf/2605.10739

18. Evaluating broad area search and classification of heavy construction activity from multisource, multitemporal satellite image sequences \- SPIE Digital Library, https://www.spiedigitallibrary.org/conference-proceedings-of-spie/13461/1346107/Evaluating-broad-area-search-and-classification-of-heavy-construction-activity/10.1117/12.3053632.full

19. GeoWATCH for Detecting Heavy Construction in Heterogeneous Time Series of Satellite Images \- arXiv, https://arxiv.org/html/2407.06337v1

20. WATCH: Wide-Area Terrestrial Change Hypercube \- CVF Open Access, https://openaccess.thecvf.com/content/WACV2024/papers/Greenwell\_WATCH\_Wide-Area\_Terrestrial\_Change\_Hypercube\_WACV\_2024\_paper.pdf

21. Spire Global Nanosatellite Constellation \- eoPortal, https://www.eoportal.org/satellite-missions/spire-global

22. Analytical review of remote sensing space programs of Russia and foreign countries, https://innoter.com/en/articles/analytical-review-of-remote-sensing-space-programs-of-russia-and-foreign-countries/

23. Inclusive energy transition \- COSMIC, https://cosmic-horizon.eu/inclusive-energy-transition

24. AI-Powered Space Exploration & Satellite Image Analysis, https://www.futurismai.com/solutions/ai-powered-space-exploration-satellite-image-analysis/

25. Indirect or Reverberating Excessive Collateral Damage in Modern IHL \- Lieber Institute, https://lieber.westpoint.edu/indirect-reverberating-excessive-collateral-damage-modern-ihl/

26. Intel office enlists AI to analyze satellite images \- Cosmic Log, https://cosmiclog.com/2021/01/06/intel-office-enlists-ai-to-analyze-satellite-images/

27. War in cities: The 'reverberating effects' of explosive weapons \- Humanitarian Law & Policy Blog, https://blogs.icrc.org/law-and-policy/2017/03/02/war-in-cities-the-reverberating-effects-of-explosive-weapons/

28. One Click from Conflict: Some Legal Considerations Related to Technology Companies Providing Digital Services in Situations of Armed Conflict | Chicago Journal of International Law, https://cjil.uchicago.edu/print-archive/one-click-conflict-some-legal-considerations-related-technology-companies-providing

29. Written evidence from International Committee of the Red Cross (ICRC) (TFP0029), https://committees.parliament.uk/writtenevidence/36625/pdf/

30. The Internal Proportionality Assessment in Cyberspace \- Lieber Institute \- West Point, https://lieber.westpoint.edu/internal-proportionality-assessment-cyberspace/

31. Overview of the Doctrine of Self Defence | LawTeacher.net, https://www.lawteacher.net/free-law-essays/international-law/overview-of-the-doctrine-of-self-defence-international-law-essay.php

32. Operation Cast Lead: The Elusive Quest for Self Defense under International Law, https://lawrecord.com/files/36\_Rutgers\_L\_Rec\_164.pdf

33. Artificial Intelligence and the “Armed Attack” Threshold in International Law \- U.S. Naval War College Digital Commons, https://digital-commons.usnwc.edu/cgi/viewcontent.cgi?article=3135\&context=ils

34. Preemptive Strikes and the Korean Nuclear Crisis: Legal and Political Limitations on the Use of Force \- UW Law Digital Commons, https://digitalcommons.law.uw.edu/cgi/viewcontent.cgi?article=1319\&context=wilj

35. International Law and the Legality of Pre-emptive Military Strikes: A Critical Analysis of the Recent US-Iran Conflict \- ResearchGate, https://www.researchgate.net/publication/410696766\_International\_Law\_and\_the\_Legality\_of\_Pre-emptive\_Military\_Strikes\_A\_Critical\_Analysis\_of\_the\_Recent\_US-Iran\_Conflict

36. Operation Epic Fury and the International Law on the Use of Force, https://carnegieendowment.org/research/2026/03/operation-epic-fury-and-the-international-law-on-the-use-of-force

37. Predictive Defense Systems and Pre-Emptive Strikes under, https://arws.cz/en/news-at-arrows/predictive-defense-systems-and-pre-emptive-strikes-under-international-law

38. The Dynamics of Japan's “Armed Attack Initiation” Doctrine and Anticipatory Self-Defense, https://www.lawfaremedia.org/article/dynamics-japans-armed-attack-initiation-doctrine-and-anticipatory-self-defense

[image1]: <data:image/png;base64,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>

[image2]: <data:image/png;base64,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>

[image3]: <data:image/png;base64,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>

[image4]: <data:image/png;base64,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>