PALANTIR: War, Surveillance, Artificial Intelligence

Part 6: The Future and Implications, Pax Technica

Chapter 14. The Future at a Crossroads

Attorney Kyungjin Kim

A. Savior or Ruler

One early morning in March 2020, a doctor in the emergency room of Newham General Hospital in East London picked up her phone. On the screen was a text message from hospital headquarters: "Oxygen ventilators assigned today: 0. No additional stock expected." She looked down the hallway. Thirty-two patients lay there, and eleven of them would not survive until the next morning without ventilators. At that moment, the same scene was being repeated across all of Britain. Someone was drinking coffee in front of a warehouse overflowing with ventilators, while someone else was holding a patient's hand with empty hands. The problem was not a shortage of ventilators. The problem was that nobody knew where anything was.

This is the background for how Palantir entered Britain's National Health Service (NHS). The British government was floundering in a sea of data chaos. Each hospital managed inventory with its own system, and those systems did not communicate with one another. There was nowhere that could show at a glance how many beds were available at which hospital, how much vaccine remained in which region, or which medical staff were infected with COVID and in isolation. Palantir's engineers connected the nationwide databases within days. Through the system they built, the British government could see national medical resource status in real time for the first time. Equipment was moved from places with surpluses to places with shortages, and vaccine distribution priorities were determined based on data. By 2024, the value of this contract reached 330 million pounds. Palantir undeniably saved lives.

(1) The Benefits of Efficiency and Security That AI and Data Integration Bring

The word efficiency gives a cold impression. But for someone dying in a hospital corridor, efficiency is another name for life. The first benefit of data integration is straightforward: it gathers scattered puzzle pieces in one place so the full picture becomes visible. As the UK example shows, the tragedy that occurs when information exists in the right places but fails to reach where it is needed is not a technical failure but a systemic one. Palantir's software served as the glue filling that failure.

On the battlefield, this effect becomes even more dramatic. We have already examined what happened in Ukraine. When hundreds of Russian tanks rolled forward, what enabled Ukrainian forces to hold was not an abundance of ammunition. It was a system that calculated in real time where the enemy was, where friendly artillery was positioned, and which weapon to strike with at what moment. Palantir's software wove satellite imagery, drone footage, field scouts' reports, and intercepted communications into one and displayed it on the commander's screen. The target identification and strike authorization process that had previously taken hours was shortened to minutes. In military terms, this is called "kill chain acceleration." Put coldly, the side that shoots faster wins. Palantir enabled Ukraine to shoot faster.

On July 31, 2025, the U.S. Army signed its largest-ever contract with Palantir. This contract, worth up to $10 billion over 10 years, consolidated 75 previously scattered individual software contracts into one. Army Chief Information Officer (CIO) Leo Garciga said in the announcement: "This enterprise agreement represents the Army's commitment to modernizing our capabilities while acting in a fiscally responsible manner. By streamlining procurement processes and using enterprise-level discounts, we are maximizing purchasing power while enhancing operational effectiveness." There is a key point here. The Army did not simply buy a new weapon. It bought the "integration" of binding 75 different systems into one. This integration is Palantir's true product.

The same logic operates in the corporate world. Airbus sources parts from around the globe to build aircraft. The number of parts in a single aircraft runs into the millions. Tracking in real time where those parts' inventories stand, when they will arrive from which supplier, and which production line is at risk of stopping is a task beyond human capacity. The Skywise system Airbus built integrates all this information and presents it to managers. When global supply chains collapsed during the COVID pandemic, the difference in recovery speed between companies with this system and those without was stark.

BP offers a similar case. Energy companies operate refineries, pipelines, storage tanks, and gas station networks. The data generated at each facility is enormous , pressure gauge readings, temperature sensor records, truck routes, fuel expiration dates. If this data is not integrated, management is effectively blind. They cannot know in advance which equipment at which refinery is about to fail, or where fuel demand is surging. Palantir's Foundry platform aggregates all this data onto a single screen. Not remediating after a problem occurs, but predicting before it happens. This is the second benefit of data integration.

Alex Karp repeats: the problems facing Western society are too complex to solve with human cognition alone. There is merit to his point. Climate change, pandemics, supply chain disruptions, terrorism, cyberattacks , all these threats are enormous, complex, and interconnected. It is impossible for a single genius to consider every variable and make the optimal decision. But machines are different. Machines simultaneously process millions of data points, find patterns invisible to humans, and calculate the optimal choices. Humans need only make the final decision.

In May 2025, the U.S. Department of Defense increased the Maven Smart System contract with Palantir by $795 million. Maven is a system that uses AI to analyze battlefield video and imagery. In the past, analysts had to manually review thousands of hours of drone footage by eye. Now AI does that work. The machine does not get tired, does not lose focus, and does not miss patterns. Analysts only need to review the critical scenes AI has filtered out. This is what "human-machine symbiosis" actually looks like. The machine does not replace the human; it amplifies human capabilities.

The third benefit of data integration is transparency. It may sound paradoxical. A company criticized as a surveillance tool touting transparency as an advantage? But Palantir's system has a built-in "audit trail" function. It records who accessed which data, when, and how it was used. The "blame avoidance" common in bureaucratic organizations becomes difficult. Because the data underlying every decision is fully preserved. Of course, the point that this transparency is a double-edged sword is something we will examine shortly.

The UK's pandemic response, Ukraine's war effort, the U.S. Army's procurement innovation, and global corporate supply chain management. All these examples converge on a single conclusion: in a complex world, the ability to integrate and analyze information is power. Palantir has made that power into a product. That their technology saves lives, leads wars to victory, and improves corporate efficiency is an undeniable fact. But as with all power, this power can produce entirely different outcomes depending on the intentions of the person wielding it. Now it is time to look at the other side.

(2) The Dystopia of Privacy Loss and a Controlled Society

In March 2025, reports emerged that the U.S. State Department had launched a program called "Catch and Revoke." The program's purpose was to use AI to analyze the social media of student visa holders, searching for statements sympathizing with terrorism. Tens of thousands of international students had their tweets, Instagram posts, and Facebook comments scanned by algorithms without their knowledge. Suppose a student posted something critical about the situation in the Gaza Strip. Could that be classified as "terrorist sympathizing"? If so, who decides? The algorithm, or a person?

This is where Palantir enters. According to reporting by the Washington Post and the New York Times, the Trump administration was expanding Palantir's technology across the federal government. At least four federal agencies , the Department of Homeland Security, the Department of Health and Human Services, the Social Security Administration, and the Internal Revenue Service , had adopted or were discussing adoption of Palantir's Foundry platform. If these platforms are connected, what happens? A single person's tax records, Social Security information, medical claims history, immigration records, and bank account information could be integrated into a single profile. A bureaucrat could peer into a citizen's entire life with a few clicks.

This is technically possible. Palantir's "ontology" is designed to do exactly this. Connecting scattered data to create a "digital twin." If this technology was used at Britain's NHS to track the location of ventilators, at the U.S. Department of Homeland Security, it could be used to track the location of people. The technology itself is neutral. But the context in which technology is used is never neutral.

On February 16, 2023, the German Federal Constitutional Court delivered a historic ruling. The Palantir-based data analysis systems used by police in Hesse and Hamburg violated the constitution. The court's logic went like this: even data that appears harmless individually, if combined without limit to create a "profile" of a person, constitutes a serious violation of the "right to informational self-determination." The judges warned: "If profiling becomes possible with a single click, citizens will not even know which of their actions are being surveilled."

In Los Angeles, this had already happened. LAPD's "Operation LASER" was a predictive policing program using Palantir's technology. The algorithm analyzed past crime data to predict areas and people where crime was likely to occur in the future. The problem was that the past crime data itself was biased. If police patrolled a particular area more frequently, more crimes were detected there. The data recorded that area as having "high crime rates." The algorithm learned from that data and classified the same area as "high risk." Police again concentrated on that area. A vicious cycle. As a result, specific racial groups and low-income neighborhoods were excessively surveilled, arrest rates climbed, and the data once again classified them as "dangerous groups." This is the "automation of discrimination by algorithm."

The relationship between ICE and Palantir is even more contentious. ICE uses Palantir's system to track undocumented immigrants. The system does not merely look at immigration records. It connects vehicle registration information, phone numbers, addresses, family relationships, employment records, and even school enrollment information to build a "network" around a person. It identifies who the undocumented immigrant's relatives are, where those relatives work, and whether other undocumented immigrants are at that workplace. This mapped information is used in large-scale enforcement operations. According to 2025 reporting, the Trump administration is pursuing plans to further expand this system by integrating federal databases.

Even within Palantir, voices of concern are emerging. The New York Times quoted anonymous Palantir employees: "There are worries about bringing this much sensitive information together in one place. The company's security practices are only as good as the people using them." Another employee worried about the impact of cooperation with the Trump administration on the company's reputation. But their concerns did not change management's decisions.

What happened in the Gaza Strip shows these concerns in their most extreme form. The "Lavender" system reportedly used by the Israeli military operated on a structure where human commanders approved AI-generated target lists. According to reports, approval took an average of just 20 seconds. Despite AI having a 10% error rate, human deliberation was skipped in favor of speed and efficiency. Whether Palantir was directly involved in this system has not been confirmed. But it is a fact that Palantir declared its official support for Israel in the Israel-Hamas war. Alex Karp wrote in his 2024 shareholder letter: "We stand with Israel. In this moment, and going forward."

The core of the problem is this. Palantir's technology offers choices. The same ontology can track the location of ventilators or track the network of an immigrant. The same kill chain acceleration can eliminate a terrorist or kill a civilian. Technology does not know purpose. Purpose is set by people. But when technology becomes too powerful, the situation can arise where technology narrows people's choices rather than people controlling technology. The logic of "we have this data, so we must use it" pushes out the question of "should we collect this data in the first place?"

As of 2025, Palantir's stock is at all-time highs. Market capitalization has exceeded $250 billion, surpassing traditional defense giants Lockheed Martin and Raytheon. President Trump said at an AI summit: "We buy a lot from Palantir." The entanglement of government and technology companies has never been deeper. When an unelected technology company assumes core national functions, where should democratic control operate?

Amnesty International warned in a 2025 report: "Palantir is failing to live up to its responsibility to protect human rights when its software is being used to facilitate serious human rights violations." Internal U.S. Army memos warned of "fundamental security" issues and vulnerabilities in joint battlefield communication systems built by Palantir and its partners. But these warnings did not prevent the signing of a $10 billion contract.

In Tolkien's novels, the palantír is "the seeing stone." Those who possess this crystal ball can see events taking place far away. But in the novels, this stone becomes not a blessing but a curse. What it shows is only a portion of the truth, and that portion can mislead the viewer. Saruman and Denethor were bewitched by what they saw through the palantír and met their doom. The warning Tolkien wrote in the 1950s is, uncannily, becoming reality 70 years later.

B. A Question for the Reader

One ordinary afternoon, a woman in a Seoul apartment was looking at her smartphone. On the screen was a notification from the Health Insurance Review and Assessment Service: "Your medical record has been updated." She cleared the notification without a second thought. At that same moment, her medical record was being entered into an insurance company's underwriting system. Also at that same moment, the keyword "cause of headaches" that she had searched for yesterday was being added to an advertising platform's profile. She knew none of this. She did not think she needed to know. But years later, when she tries to buy insurance, or goes to a job interview, or applies for a loan, these data points could influence her fate. Without her ever knowing.

This is the world we live in.

(1) In the Era of Data Power, What Must We Watch?

Palantir is a mirror. Looking into this company, you can see how the data power of our era operates. But we cannot just stare at the mirror. The question we must now ask is this: in the era of data power, what must we as citizens watch?

The first thing to watch is "the flow of decisions." Not the data itself. Data is merely raw material. What matters is how that data is combined, what conclusions it produces, and what actions those conclusions lead to. When LAPD's predictive policing system classified a particular area as "high risk," on what basis was that classification made? Was there a procedure to challenge that classification? When the classification was wrong, who was held accountable?

Bruce Schneier and Nathan Sanders wrote in their 2025 book, Rewiring Democracy: "AI is fundamentally a power-amplifying technology. It executes one person's command at greater speed, larger scale, broader scope, and more sophisticated manner than any human individual could." The question is to whom this amplified power goes. To citizens, or to the bureaucrats and corporations managing citizens?

The second thing to watch is "the fiction of consent." Every day we "consent" to dozens of apps and services. But nobody reads the terms. The terms are written to be difficult to read, and even if you do read them, there is often no option to refuse. Research frequently finds that data is still collected even when users have declined tracking. Consent has become a mere formality. Real control does not rest with the user.

Research on Facebook (Meta) reveals an interesting phenomenon. On the front stage, the platform manages user trust through visible measures like "privacy settings" and "real-name policies." But backstage, much more significant data flows move almost invisibly through recommendation algorithms and ad targeting systems. In theatrical terms, it is a separation of "front stage" and "back stage." What we need to watch is not the show on the front stage, but what is actually happening backstage.

The third thing to watch is "the gaps in regulation." Blockchain, the metaverse, generative AI, cross-border data flows , in these new technological environments, gaps emerge that existing laws and institutions fail to adequately address. Comparative law research on metaverse platforms shows that countries still take fragmented approaches to consumer protection, jurisdiction, privacy, and liability governance. These gaps lead to the expansion of "private regulation," where companies set their own rules. Who makes the rules, and who watches the rule-makers? This is the essential question.

The fourth thing to watch is "the use of data on vulnerable populations." Research on children's app marketplaces reveals a shocking reality. Platforms constitute child users as "commodified audiences." Children's clicks, viewing times, and preferences are collected as data and sold to advertisers. As data in education, welfare, and healthcare becomes increasingly quantified and serves as material for algorithmic evaluation, the risk grows that the lives and possibilities of vulnerable populations will be narrowed by data models.

According to a 2025 Carnegie Foundation survey, 55% of California residents responded that they were "very concerned" that AI-generated content would deepen political violence and polarization. At the same time, citizens were divided on whether AI could make them "more informed citizens." This data suggests an important point: anxiety about technology is high, but how to translate that anxiety into action has not yet been agreed upon.

The fifth thing to watch is "apathy itself." Many people press the "agree" button without reading privacy policies. This is not individual laziness. It is the result of structural fatigue caused by a data environment that is difficult to understand and designed unilaterally. If this situation persists, data power expands without meaningful resistance. Citizens' democratic agency gradually shrinks. That is why efforts are needed to reposition data issues, through education, media literacy, and public discourse, not as the domain of experts but as part of citizens' common sense and political imagination.

Ultimately, the central question of the data power era goes beyond "Are we being surveilled?" The more important question is this: "What, and how, shall we watch together?"

(2) Beyond Palantir: Imagining the Next Generation of Data Governance

Moving beyond Palantir does not mean eliminating Palantir. It means telling a bigger story. Palantir is one case showing the extremes of the current paradigm. How quickly data integration can make a state's decisions, how a platform becomes infrastructure, and what ethical burdens that infrastructure carries , all of this can be seen through the lens of Palantir.

What we must now imagine is the next generation of data governance.

First, "democratic data agency" must be institutionalized. This means a structure where users are not mere consent providers but subjects who participate in discussing and deciding data policies and usage directions. New institutional imagination is needed: data cooperatives, citizen panels, participatory data governance bodies. In February 2025, California announced a new deliberative democracy program , an experiment using AI to analyze and integrate the opinions of thousands of citizens and reflect them in policy decisions. It is still in its early stages, but it shows the possibility that technology can be used to expand citizen participation.

Second, "responsible governance" principles must be built into the design phase of technological infrastructure. The European food and nutrition data platform case offers a good example. This platform applied design principles that holistically considered data types and technologies, ownership and intellectual property, privacy and security, and ethical governance structures. It demonstrates that a data platform can be not merely a means of increasing efficiency, but a space where individual and collective privacy, power relationships, and accountability are designed together. Data governance is not only a matter of law and policy. It is also a matter of "design ethics" embedded in platform architecture and operating norms.

Third, moving beyond closed, proprietary models, we must explore alternative platform ecosystems that strengthen public interest. California's proposed "CalCompute" initiative envisions a structure where open-source AI systems operated on public servers are governed by public institutions. A public interest data archive model openly accumulates and analyzes materials for monitoring and holding platforms accountable , contracts, court documents, impact assessments, patents, and civil society reports , providing a foundation for researchers, journalists, and activists to demand platform accountability.

Fourth, the framework for international data governance must be reconfigured around human rights and the public good, not only competition and security. Currently, platform governance centered on the United States and China emphasizes algorithmic sovereignty and data localization in the name of security and economic competitiveness. But there are also significant gaps in governance and accountability for data flowing across borders. The next generation of data governance requires redefining cyberspace not as a "deregulated, deterritorialized zone" but as a space where international human rights standards and accountability norms apply.

Fifth, data sensitivity and imagination must be cultivated at the educational and cultural level. Studies analyzing surveillance capitalism warn that the current data regime is permeating daily life with little social resistance and forming a new social order. To change this structure, data and algorithmic issues must be made topics that citizens, students, and children can discuss together , not the exclusive domain of technologists or policy experts. When learning about and discussing data rights, privacy, algorithmic bias, and platform accountability becomes routine in schools, homes, and communities, the next generation will be able to imagine the post-Palantir data order more critically and creatively.

In a September 2025 report, the OECD recommended: "Government AI adoption is growing, but many initiatives remain at the pilot stage. Successful AI adoption requires seven key elements: governance, data, digital infrastructure, skills, investment, procurement, and partnerships with non-government actors. A whole-of-government, coordinated approach to AI transformation that is user-centric, meaningfully engages stakeholders and the public, and establishes clear and proportionate guardrails will enable managing risks while remaining agile and adaptable to future changes."

The final question returns to the reader. Whether to view the era of AI and data integration solely as a narrative of "salvation" or solely as a narrative of "domination" is not a predetermined fate. The boundary between the two will shift according to what governance structures we design, what power we monitor, and what language of participation and resistance we create.

Alex Karp says: "Technology is not neutral." He is right. Technology reflects the worldview of those who created it. If so, we must demand technology that reflects not the worldview designed by a handful of genius engineers or philosopher-CEOs, but democratic values agreed upon by ourselves.

If Palantir's software is used to catch terrorists, it must also be usable to monitor police abuse of power. Just as technology serves those in power, it must be designed to serve as a shield protecting the vulnerable.

In Tolkien's novels, the palantíri are ultimately thrown into the sea. Because it was the only way to escape their temptation and danger. But the real-world Palantir cannot be thrown into the sea. The era of data integration and AI has already arrived, and there is no going back. What we can do is participate in deciding who uses this powerful tool, how, and for what purpose.

The Technological Republic has arrived. But whether the sovereign of that republic is technology or citizens has not yet been decided. It is not Palantir standing at the crossroads. It is us. Remember that even the most powerful software in the world cannot execute a single line of code without human will. The moment our vigilance and participation ceases, on the other side, someone's finger will be poised over the Enter key.

The Night Outside the Window

As I write the final sentences of this book, Seoul's night is quiet. Lights twinkle outside the window. The people sleeping beneath those lights have likely never heard the name Palantir tonight. They will not know where their data flows or who is watching it.

Throughout the writing of this book, one question never left my mind. Are we spectators in this game, or pawns?

Palantir's technology was born in the embrace of the United States, a military superpower. They created this tool for their own national security, and they selectively permit it to allies. Korea is one of those allies.

But behind the comfort that the word "ally" provides lies an uncomfortable truth. We did not create this technology. We are in the position of borrowers.

There is a term called data sovereignty. The principle that a nation's data should be controlled by that nation. But in reality, this principle wavers before the logic of power. Wherever Palantir's servers are located, whatever the people who designed those algorithms are thinking, we cannot know. There is no way to verify how our information is processed inside the black box.

This is the source of regret.

As a lawyer, as a former member of the National Assembly, as someone who researches artificial intelligence, I believe in the power of technology. But when that power is concentrated in the hands of a few, and when those few are people from another country, I have not found the answer to how democracy should function.

Palantir says it defends Western democracy. Whether we are included within that defense fence, or whether we are on the outside looking in, I am not sure.

In January 2026, President Trump declared he would "get" Greenland, an autonomous territory of Denmark, "one way or another."

If not the nice way, then the hard way. To use his expression, it was "the hard way." Denmark is a founding member of NATO. It has been America's ally for over 70 years. Yet the president of the United States was publicly threatening the territory of that ally. Vice President Vance visited the U.S. military base in Greenland without an invitation and gave a speech saying "Denmark has let down the people of Greenland." The wife of a White House aide posted a map of Greenland covered in American flags on social media and wrote "SOON."

Danish Prime Minister Mette Frederiksen's response was urgent: "If the United States takes military action on Greenland, it will be the end of NATO." Denmark's intelligence service, for the first time in history, listed the United States alongside Russia and China on its national security threat list. In Copenhagen and Nuuk, citizens took to the streets carrying placards reading "Yankee go home."

A country that styles itself the defender of Western democracy is threatening a core ally of Western democracy. A country that invoked international law to criticize China and Russia is now ignoring the same international law and coveting an ally's territory. We are witnessing the moment when the slogan "peace through strength" is being distorted into "submission through strength."

One cannot help but ask: if even a founding NATO member receives this treatment, what about Korea?

We do not belong to the "West" they speak of. When our data passes through their servers, when our military intelligence is analyzed by their algorithms, are we partners, or customers? Or perhaps, products?

No matter how superb Palantir's technology may be, we cannot control the intentions of those who own that technology. The Greenland crisis showed with brutal clarity that today's ally can become tomorrow's coercer. That data sovereignty is not an empty slogan but a matter of survival , I only now feel this in my bones.

I want to ask the reader who closes this book: What are we giving up in exchange for convenience? And who should judge whether that trade is fair?