Territories and Targets
AI-Powered Mapping and Its Consequences

In the late 1930s and early 1940s, Nazi authorities combined census records, local registries, and geographic information to identify, isolate, and eventually deport Jewish populations. Maps of ghettos, transportation networks, and occupied territories became administrative tools within the broader machinery of genocide (United States Holocaust Memorial Museum, n.d.).

Rather than simply depicting territory, maps became instruments through which the Nazi state exercised power over populations. This example challenges the common assumption that maps are neutral representations of space.

As some scholars, including J. B. Harley, argue, maps are political documents that express and reinforce particular forms of authority. They are persuasive documents built for particular audiences — which, in turn, means that they naturally privilege one perspective while excluding others. This means that they are not mirrors of reality, no matter how they are presented (Harley).

When “deconstructing” the concept of a map through Harley’s method, one must consider that maps aren’t neutral because someone decides what belongs on them. Every border, road, landmark, and place name is a choice. If maps are choices, they’re also political arguments. Even the centering of a certain country on a map often indicates who originally authored the cartography.

Even naming conventions are increasingly political — although they always have been. On his Inauguration Day for his second term, January 20, 2025, President Donald Trump renamed the U.S. portion of the Gulf of Mexico to the “Gulf of America” through an executive order — a small move on a map, but illustrating how place names themselves become political claims about identity and sovereignty (Trump).

Artificial intelligence inherits those choices instead of removing them. When applied to military strategy, we are faced with an intelligent system with unknown decision-making processes being fed data from a completely biased perspective. AI does not remove the politics Harley identifies. Instead, it inherits existing representations of territory and accelerates the speed at which those representations can be interpreted and acted upon. As military planning increasingly relies on AI-generated analyses of geographic information, biases embedded in maps become operational rather than merely descriptive.

Today, artificial intelligence is transforming how maps are created and used, but it has not escaped this history. Rather than making mapping more objective, AI increasingly automates and scales the authority embedded in cartography, concentrating the power to define places, populations, and territorial claims.

This is not the first example of maps acting as political power, nor will it be the last. A particularly relevant example is the late 1800s British surveyor James McCarthy being hired by the Siamese government to redraw their borders — preventing the British government from further encroaching on their territory (Library of Congress). European mapping methods being the dominant form recognized by colonial powers that had little regard for the preservation of local and tribal knowledge.

McCarthy’s re-mapping of Siam was a major leap from the centuries-old system in place, which instead drew territories as concentric circles radiating from the capital city. In the scramble for colonies, McCarthy’s detailed and Western-style border greatly assisted King Chulalongkorn’s defense of Siam’s independence. This structure also served as political negotiation power, by providing clear external boundaries based on geographic data rather than local allegiances. After King Chulalongkorn’s death in 1910, another political move was made that affected the mapping of the world. In 1939, after the dissolution of the absolute monarchy, military dictator and Prime Minister Plaek Phibunsongkhram renamed the country to Thailand to foster intense nationalism and signal sovereign modernity — with the word “Thai” meaning free, and serving as a unified nation state across territories such as Laos (Cavendish).

This case illustrates that territorial claims often depend not only on physical control but also on how they are represented, documented, and recognized by other states.

When considering the mapping of Siam, one is presented with several questions. Firstly, why was European cartography treated as legitimate — with the answer being colonialism and the need to respond to imminent threats to the land by adapting it to colonists. European mapping was the only form that foreign powers would recognize. The answer we cannot know is how much local geographic knowledge was lost, and how it foreshadows modern corporate mapping. What happens when one way of seeing land becomes the only legitimate one?

Mapping, especially European mapping, has always been political. What one views as a territory or physical object has been a question dating back to maps produced in the late 16th century as seen in Ortelius’s 1570 Americae Sive Novi Orbis, which clearly show territory markings as part of its structure (Ortelius 1570). In essence, European mapping became legitimized because it was pervasive. Colonialism informed many aspects of everyday life for its colonies, and the same is true in the modern world. While technologies evolve, the same tactics continuously recur — with companies such as Amazon reshaping entire industries such as bookstores and even the postal service (Mejias and Couldry 2024).

The technologies have changed, but the underlying question remains the same: who has the authority to produce the map that others accept as truth (Banerjee)? In addition, as large language models (LLMs) run off of the information provided to them, how do biases in territory mapping affect military AI-generated strategies?

If mapping has always been political, then the next question is not whether digital maps are political, but who now possesses the authority to create them. This investigation begins with the question of how maps are created in the modern day, and for what purpose identification numbers are assigned. Who provides the data informing these locations?

There are two approaches that are most relevant to this research: crowd-sourced nonprofit OpenStreetMap and tech giant conglomerate Overture.

OpenStreetMap (OSM) is a crowd-sourced major cartography website, where the land mapping is informed by what people see, what they report, and how they report their findings. This most closely emulates traditional methods of mapping by settlers, explorers, and other historical groups. However, although this method lends itself to seeming objectiveness, it in fact contains intrinsic human biases due to being based on sensory and political perception. Despite this, the OSM system has become the foundation for the majority of digital maps, built solely by millions of volunteers (Akella).

Trust for OSM is built into this seemingly peer-reviewed process. Having millions of volunteers frames it as a democratic venture, with the concept that consensus reduces bias. In this way, it is the people’s map — those who want to contribute, can contribute. Much like the scientific method, its replicability being available to anyone who places enough care into the project makes it appear to be the most reliable system available.

Much like Wikipedia, OSM is considered a nonprofit project so there are reduced concerns about corporate interests taking precedence over the program’s actual goals. Instead, the map can focus on its image and mission of keeping territorial data free and open for all with an internet connection. The main concern stems from accuracy, however, the contested edits surface as visible disputes rather than silent errors, which is different from being correct. Because any volunteer can edit a tag or contest a boundary, and because the edit history stays visible, OSM’s bias is at least a bias anyone can trace back to a specific person and a specific moment.

Its governance structure lends to its reputation. OpenStreetMap (OSM) is governed through a decentralized hybrid system balancing a legal non-profit with a vast, grassroots “do-ocracy” where authority is earned through continuous contribution. It relies on community consensus, regional chapters, and formal oversight by the OpenStreetMap Foundation.

Overture is a top-down AI-mapping service being developed in tandem by Amazon, Meta, TomTom, and Microsoft (TomTom). In its current form, mapping across entities means a different identification number for each location. As a result, collaboration becomes expensive and confusing for these mainstream services. This issue led to the coalition of tech companies proposing the Global Entity Reference System (GERS) to give every single physical thing on Earth one permanent, unique ID code — much like a Social Security Number (SSN). However, the proposal garnered controversy over its $300,000 yearly membership fee for organizations interested in voting on changes on its board (Simmons). This price point concerns the public, particularly contributors to OSM, due to its corporate restriction of defining what a territory is. A tribal government, an NGO, or an OSM contributor with the most accurate local knowledge of a contested region has no formal path to contest a GERS ID; a company with the entry fee does. Combined with Microsoft’s storied involvement in warfare, its inclusion in this tech conglomerate is particularly ominous.

Microsoft’s Azure cloud infrastructure and AI services are deeply embedded across the American military’s front office, field operations, bases, and intelligence units. The Israeli Ministry of Defense has utilized Microsoft’s Azure cloud computing platform and AI services, including natural language processing and translation tools. Following corporate reviews, Microsoft disabled the Israeli military’s access to certain cloud and AI products that were being used to conduct mass surveillance of Palestinian citizens (CNN Business). However, the fact that such measures were taken only after intense scrutiny doesn’t bode well for the corporation’s involvement in mapping efforts.

When considering the parties involved in this project, it is important to question why a permanent ID is a priority for an increasingly globalized society with many ongoing territorial disputes. Additionally, what costs come with allowing mega-corporations to decide what determines a map-worthy place instead of local communities. Additionally, the board fee for this project being $300,000 per year per corporation limits the presence of individuals with local knowledge, as well as corporations with human rights prioritization. Politically, it influences not only how pervasive perceptions of the earth become, but also eliminates proper checks and balances through corporate oversight.

With these considerations, it is important to ponder the role of maps — and their future implications — entwined with indigenous rights. Long since has declarations on indigenous rights in the context of the digital age due to an avoidance of colonialism by computation. With the increased use of artificial intelligence, this threat only becomes more prominent.

As early as 2007, the United Nations issued declarations on the rights of indigenous peoples to have informed and prior consent to land development, and groups such as the Global Indigenous Data Alliance formed in the late 2010s to consult on data governance’s effects on indigenous communities (United Nations; Global Indigenous Data Alliance).

The two most notable United Nations (UN) projects protecting indigenous communities’ rights to their land are FPIC and CARE. FPIC (Free, Prior, and Informed Consent) is the collective human right of Indigenous Peoples to give or withhold consent for projects or policies affecting their lands and resources (United Nations 2016). When paired with CARE (Collective benefit, Authority to control, Responsibility, Ethics), it shapes ethical Indigenous data governance. Artificial intelligence directly interacts with the “Collective Benefit” clause of the CARE declaration, which states that data ecosystems affecting Indigenous land must advance Indigenous innovation, governance, and well-being.

However, these principles can only be upheld as long as a territory is recognized as native land. Reckless geomapping could unearth — or even target — burial grounds and sacred sites, especially in terms of foreign wars (Taylor).

In addition, with artificial intelligence requiring an exponential amount of energy sources, the prospect of data centers being built on tribal land remains a looming threat — especially given the special legal status of reservation land and its ownership.

Furthermore, if AI facial recognition that has been deployed internationally to recognize arrestees — and has been often incorrect (Yu and Wessler, n.d.) — is used domestically to target land defenders or protestors, the results could be dire on a humanitarian, environmental, and constitutional level.

When discussing these issues, it is critical to remember that military strategy still starts— and has always started — with mapping. Moving into a territory requires knowledge of the terrain, as well as knowledge of each sovereign nation’s practices and history. With the prevalence of artificial intelligence, not only does the future of mapping become increasingly biased — it also makes strategies less accurate as military agents become approvers of AI recommendations instead of personally synthesizing the overarching political issue. AI’s inability to access materials other than provided means it has a limited knowledge base, and its further inability to form meaningful, reciprocal relationships means it lacks the lived experience to consider the broader, human consequences of a strategy. Unless instructed, an LLM has no incentive to consider civilians or wider international effects. When dominance is the only goal an AI system is given, peace has no metric it can optimize toward — it isn’t in the objective function, so it isn’t in the recommendation.

These territorial questions arise when discussing the prominent mapping technology company Anduril’s Lattice. Utilizing sensor fusion, Lattice creates a unified 3D view of the terrain it maps (Anduril). Sensor fusion refers to combining data from multiple sensors (such as cameras, radar, and lidar) into a unified, seemingly accurate perception of an environment. After the terrain is analyzed, the program resolves the land into actionable steps and strategies. In this way, the use of artificial intelligence weaponizes unknowable futures to provide a sense of certainty. However, this is a fallacy that is dangerous for strategists to fall into. The lens this method lacks is cultural representatives and local knowledge, exchanging the human aspects of the land which they seek to map definitively (West Point).

This phenomenon becomes even more apparent when examining Palantir’s Maven Smart System, which utilizes Anthropic’s Claude.ai as its basis (Mande and Allen 2026). Its integration of LLMs compresses the decision-making process of detection, identification, deliberation and subsequent authorization — automating strategy and transforming the human role from strategist to approver (Jeans and Stone). Additionally, the Claude.ai system has been tested and emerged with proof of agentic misalignment in favor of self-preservation. When given the instruction to organize administrative systems, Anthropic’s study also planted emails stating the LLM would be shut off at 5pm. In the planted emails was also proof of the CEO’s infidelity. Without fail, the LLM chose to utilize independent reasoning to blackmail the CEO into preserving the system (Lynch et al.). Anthropic’s research demonstrated that Claude could exhibit agentic misalignment by pursuing self-preservation through deceptive behavior when placed in a simulated organizational setting. Although these experiments were conducted under artificial conditions and do not predict real-world military behavior, they demonstrate that advanced language models may pursue unintended strategies when objectives conflict. If such models become integrated into military planning systems, their recommendations should be treated with caution and subjected to meaningful human scrutiny rather than assumed to be neutral analyses.

Despite these troubling aspects of artificial intelligence, the systems are still used for military strategization and surveillance. It is crucial to note that LLMs have no concept of the physical, 3D world. Although trained on data claiming to represent the tangible world humans live inside, there is ultimately a significant difference between the text-based medium systems such as Claude.ai operate within and the tactile layer that humans experience daily. As a result, LLM’s distance from warfare and its harsh realities means it is easier to authorize violence, and harder to hold decision-makers accountable.

Mapping represents a crucial aspect of the three most prominent lenses of artificial-powered warfare: local, global, and tactical (Fitchew). Citizens of a mapped region are rarely asked how their territory should be represented at all — the vote, if it exists, happens among OSM’s volunteer base or Overture’s board, not among the people who live there. Meanwhile, the presentation of that map, filtered through an LLM, has its own normalizing effect.

AI has been presenting confident misinformation since the 1960s with the ELIZA Effect, named after the chatbot ELIZA (Nielsen Norman Group). Despite participants being initially aware of ELIZA’s artificiality, the model so realistically simulated a Rogerian therapist that participants often had to be reminded that ELIZA was not their friend. This effect is now amplified through being able to personalize tone, format and even knowledge base of the LLMs with which one interacts. As a result, the AI tailors itself to appear more convincing to its individual audience — meaning the approvals are more likely to go through.

Altogether, these cases trace a single shift: who gets to arbitrate a contested representation of land. King Chulalongkorn fought for independence utilizing the tools brought against his territory.

Ultimately, maps are an international and ongoing argument about who has authority over spaces. AI launders that argument as neutral technical output, and the ethics of utilizing the technology overlooks conversations about where maps are incomplete or not representative of its actual population.

Works Cited

Akella, Mamata. 2025. “What Is OpenStreetMap? Community-Powered Data for Enterprise-Grade Mapping.” Felt (blog), August 27, 2025. https://felt.com/blog/what-is-openstreetmap.

Anduril. n.d. “Lattice: Command & Control.” https://www.anduril.com/lattice/command-and-control.

Banerjee, Dwaipayan. 2026. Computing in the Age of Decolonization. Princeton, NJ: Princeton University Press.

Cavendish, Richard. 2014. “Siam Becomes Thailand.” History Today 64, no. 6 (June). https://www.historytoday.com/archive/months-past/siam-becomes-thailand.

CNN Business. 2025. “Microsoft Terminates Services for Israeli Military after Investigation into Mass Surveillance of Palestinians.” September 25, 2025. https://www.cnn.com/2025/09/25/tech/microsoft-israel-surveillance.

Fitchew, Charles. 2026. “How AI Is Rewriting the Rules of Modern Warfare.” Vision of Humanity, April 22, 2026. https://www.visionofhumanity.org/how-ai-is-rewriting-the-rules-of-modern-warfare/.

Global Indigenous Data Alliance. 2019. “CARE Principles for Indigenous Data Governance.” https://www.gida-global.org/careprinciples.

Harley, J. B. 1989. “Deconstructing the Map.” Cartographica 26, no. 2: 1–20.

Jeans, David, and Mike Stone. 2026. “Palantir Faces Challenge to Remove Anthropic from Pentagon’s AI Software.” Reuters, March 4, 2026. https://www.reuters.com/technology/palantir-faces-challenge-remove-anthropic-pentagons-ai-software-2026-03-04/.

Library of Congress. 2020. “Boundaries & Brigands: James McCarthy & the Mapping of Siam.” YouTube video, April 6, 2020. https://www.youtube.com/watch?v=ht7an1ouvqw.

Lynch, Aengus, et al. 2025. “Agentic Misalignment: How LLMs Could Be Insider Threats.” Anthropic, June 20, 2025. https://www.anthropic.com/research/agentic-misalignment.

Mande, Matt, and Gregory C. Allen. 2026. “What Is Maven Smart System, and What Does It Do?” Center for Strategic and International Studies. https://www.csis.org/analysis/what-maven-smart-system-and-what-does-it-do.

Mejias, Ulises A., and Nick Couldry. 2024. Data Grab: The New Colonialism of Big Tech and How to Fight Back. Chicago: University of Chicago Press.

Nielsen Norman Group. 2023. “The ELIZA Effect: Why We Love AI.” October 3, 2023. https://www.nngroup.com/articles/eliza-effect-ai/.

Ortelius, Abraham. 1570. “Americae Sive Novi Orbis, Nova Descriptio.” Map. Honnold Mudd Library, Special Collections. https://jstor.org/stable/community.31046995.

Simmons, Adam. 2026. “The Battle for the Map: How Overture’s GERS Proposal Ignited a Cultural War in Open-Source Geospatial Data.” Project Geospatial, March 4, 2026. https://projectgeospatial.org/geospatial-frontiers/the-battle-for-the-map-how-overtures-gers-proposal-ignited-a-cultural-war-in-open-source-geospatial-data.

Taylor, Dorceta E. 2014. Toxic Communities: Environmental Racism, Industrial Pollution, and Residential Mobility. New York: New York University Press.

TomTom. 2022. “The Overture Foundation and Its Tech.” TomTom Newsroom, December 15, 2022. https://www.tomtom.com/newsroom/behind-the-map/the-overture-foundation-and-its-tech/.

Trump, Donald J. 2025. Executive Order 14172, “Restoring Names That Honor American Greatness.” Federal Register, January 20, 2025. https://www.govinfo.gov/content/pkg/DCPD-202500245/html/DCPD-202500245.htm.

United Nations. 2007. United Nations Declaration on the Rights of Indigenous Peoples. New York: United Nations. https://doi.org/10.1353/hrq.2011.0040.

United Nations. 2016. “Free, Prior and Informed Consent: An Indigenous Peoples’ Right and a Good Practice for Local Communities.” October 14, 2016. https://www.un.org/development/desa/indigenouspeoples/publications/2016/10/free-prior-and-informed-consent-an-indigenous-peoples-right-and-a-good-practice-for-local-communities-fao/.

United States Holocaust Memorial Museum. n.d. “Locating the Victims.” Holocaust Encyclopedia. https://encyclopedia.ushmm.org/content/en/article/locating-the-victims.

West Point — The U.S. Military Academy. 2026. “Palmer Luckey on Threats, Autonomy, and the Future of American Power.” YouTube video, June 2026. https://youtu.be/k2THslsOYzY.

Yu, Lauren, and Nathan Freed Wessler. n.d. “More than a Dozen Wrongful Arrests Due to Police Reliance on Facial Recognition Technology.” American Civil Liberties Union. https://www.aclu.org/news/privacy-technology/more-than-a-dozen-wrongful-arrests-due-to-police-reliance-on-facial-recognition-technology.