PALANTIR: War, Surveillance, Artificial Intelligence

Part 6: The Future and Implications, Pax Technica

Chapter 13. Korea's Choice: Proposals for an AI G3 Leap

Attorney Kyungjin Kim

A. The Need for Korean Defense AI

(1) Responding to North Korean Threats and Regional Arms Races

In the fall of 2024, a massive A-shaped structure rose inside the KINTEX exhibition hall in Goyang, Gyeonggi Province. It was a Hanwha Group exhibition pavilion. Inside were AI-piloted unmanned combat aircraft models, hybrid weapons combining drones and loitering munitions, and smart warships that halved crew requirements. The theme of the Seoul International Aerospace and Defense Exhibition ADEX 2025 was "The Future Battlefield Created by AI." A Hanwha representative explained to visitors: "Now it is data, not pilots, that commands the battle."

At the same time, on the other side of the Demilitarized Zone, a different kind of innovation was underway.

North Korea launched ballistic missiles dozens of times in 2024 alone. Among them were weapons claiming to be hypersonic glide vehicles. Nuclear warhead miniaturization technology had become more sophisticated. Simultaneously, North Korea's cyberattack capabilities had reached world-class levels. According to National Intelligence Service estimates, North Korea steals hundreds of millions of dollars in cryptocurrency annually, with a significant portion of that funding going to nuclear and missile programs. It is a strategy of offsetting conventional force inferiority with asymmetric capabilities.

The problem is not North Korea alone.

Northeast Asia is the region where the arms race is accelerating fastest in the world. China has openly declared the goal of acquiring the capability to forcefully unify Taiwan by 2027. The PLA Navy has already surpassed the United States in ship count. Japan reinterpreted its constitution to raise defense spending to 2% of GDP. Russia's invasion of Ukraine left a dangerous precedent that changing the status quo by force is possible in Northeast Asia as well.

The military personnel available to Korea for responding to all these threats is shrinking. As of 2025, annual births number approximately 250,000. This figure has declined to one-quarter of the one million recorded in the 1970s. According to Ministry of National Defense estimates, by 2040 the pool of military service-eligible personnel will bottom out. The demographic cliff is not merely a social issue; it is the essence of a security crisis.

The solution the Ministry of National Defense has produced is "Defense Innovation 4.0."

The core of this plan, announced in 2023, is replacing manpower with technology. Manned-unmanned teaming combat systems where AI-based drones and robots conduct operations alongside soldiers; intelligent C4I systems where AI analyzes battlefield situations and supports commanders' decisions; Joint All-Domain Command and Control (JADC2) systems where satellites and drones collect real-time intelligence linked to strike systems. All of this is contained in the roadmap.

In April 2024, the Defense AI Center was established under the Ministry of National Defense. In January 2025, this organization was expanded and reorganized into the Defense AI Technology Research Institute. Its budget is approximately 15 billion won. Each service branch , Army, Navy, and Air Force , is pursuing more than 30 AI-based force transformation projects. Automated surveillance video analysis systems, cloud systems using AI semiconductors, and supercomputers for combat staff are being introduced.

But an enormous gap exists between reality and goals.

Korea's defense R&D budget is about 5% of total defense spending. The United States invests over 15%. An even more serious problem is data. AI requires large volumes of high-quality data to learn. But the Korean military has not established clear governance on what data is, where it exists, or how to use it. The Army, Navy, and Air Force each operate their own networks, and those networks are not standardized. The "silo" problem U.S. intelligence agencies experienced during 9/11 is being repeated inside the Korean military 20 years later.

On the Ukraine battlefield, Palantir's software integrated satellite imagery, drone reconnaissance, and human intelligence in real time, completing a kill chain in 20 minutes. According to Ministry of National Defense plans, it will take Korea 10 to 15 years to implement such a system. In the meantime, threats from North Korea and neighboring countries will not wait.

One defense expert assessed it this way: "We are still drawing the future on paper. But other countries are already testing that future on the battlefield."

(2) Public-Private Partnership (PPP) Lessons from the Palantir Model

In 2016, Palantir sued the U.S. Army.

Under federal procurement regulations, the government must first assess whether commercial software can meet requirements. But the Army was skipping this procedure and awarding contracts only to traditional defense contractors. Palantir's argument was straightforward: "If our product is better and cheaper, why aren't we even given a chance to compete?" The court ruled in Palantir's favor. The ruling included this statement: "The Department of Defense's failure to seriously consider commercial solutions is unlawful."

This ruling changed the landscape of the U.S. defense industry.

It opened the door for Silicon Valley startups to enter a market monopolized for more than 70 years by traditional defense contractors like Lockheed Martin, Raytheon, and Northrop Grumman. After that, companies like Anduril, SpaceX, and Shield AI won defense contracts. The $10 billion contract Palantir signed with the U.S. Army in 2025 was the pinnacle of this trend , a project to integrate 75 individual software systems into a single platform.

What can Korea learn from this model?

First, institutional mechanisms are needed to channel civilian technology capabilities into defense. Palantir's core technology, ontology, was originally developed for fraud detection at PayPal. Airbnb's data processing technology, Tesla's autonomous driving algorithms, and Starlink's satellite communications network were all born in the civilian sector and expanded into the military domain. Korea also has world-class semiconductor companies, AI startups, and software talent. The problem is that access points for these players to enter the defense sector are blocked.

Korea's defense procurement system remains closed. The Defense Acquisition Program Administration's complex regulations, lengthy project cycles, and cost estimation methods are unsuitable for fast-moving technology companies. One industry representative at ADEX 2025 said: "We create technology where generations change every three years. But government project cycles are 10 years. If we don't close this gap, we cannot maintain technological superiority."

Second, the "cost-plus" contracting practice must be broken. Cost-plus means paying the contractor their costs plus a set profit margin. Under this structure, contractors have no incentive to reduce costs. In fact, inflating costs increases profits. Palantir directly challenged this practice. "We will deliver results at a fixed price. Even if costs exceed the estimate, we bear the burden." This fixed-price contracting approach improved the Defense Department's budget efficiency.

Third, a shift in mindset regarding software capability is needed. Traditionally, defense was the domain of steel, ammunition, and engines. How many tanks and fighters one had was the measure of military power. But the Ukraine war upended this equation. Russia's military, once rated the world's second-strongest, suffered massive losses from Ukraine's software-based kill chain. Hardware superiority could not offset software inferiority.

Korea's defense industry senses this change. At ADEX 2025, Hanwha Aerospace unveiled the "Chunmoo 3.0" system, where AI enables real-time collaboration between reconnaissance drones and loitering munitions for target engagement. LIG Nex1 introduced an AI-based Battle Management System (BMS 2.0) integrating the entire battlefield. Hanwha Systems' Smart Battleship uses AI to reduce crew from 60 to the mid-30s. The four major defense companies' order backlog for Q3 2025 reached 91 trillion won, up 21.6% year-over-year, and an annual backlog exceeding 100 trillion won is all but certain.

But this alone is insufficient.

The true core of the Palantir model is not technology but governance. Trust between the private sector and government, free flow of data, a culture that tolerates failure, and a performance-based reward system. These elements must combine for innovation to be possible. The Presidential Committee on National AI Strategy, launched in September 2024, selected "Defense and Security" as one of its 12 strategic areas. The committee's defense and security division head is the director of the AI and Informatization Research Division at the Korea Institute for Defense Analyses. The government has begun addressing defense AI at the national strategy level.

One defense analyst offered this forecast: "Korea's defense exports are evolving from hardware-centric to system-centric. The next step is software. If we can export a data platform alongside weapons systems, as Palantir does, K-defense will enter an entirely new dimension."

Defense exports of $20 billion by 2030 , that is the government's target. To achieve it, tanks and self-propelled guns alone are not enough. What is needed is the software connecting them into one, and the public-private partnership ecosystem to create that software.

B. Data Sovereignty and National Security Strategy

(1) Risks of Dependence on Foreign Platforms and Building Indigenous Capabilities

In early 2025, the Personal Information Protection Commission released the results of its review of the Chinese AI service DeepSeek.

The results were alarming. DeepSeek was storing Korean users' data on servers in China. Under China's National Intelligence Law, Chinese companies are obligated to cooperate with the government's intelligence collection. In theory, the Chinese government could access Korean user data held by DeepSeek. The Personal Information Protection Commission issued a corrective recommendation.

This case posed a single question: when data crosses borders, where does sovereignty reside?

What Palantir sells to the U.S. government is not just software. It is control over the data that software processes. Intelligence agency data entered into the Gotham platform, corporate data integrated into Foundry, real-time battlefield data analyzed by AIP , all of it passes through Palantir's system. Who operates the system is the question of who controls the information.

Korean government agencies and companies are increasingly dependent on foreign cloud platforms. Amazon Web Services (AWS), Microsoft Azure, and Google Cloud dominate the domestic market. As of 2024, foreign companies' share of the domestic cloud market exceeds 70%. Government ministries and public institutions face the same challenge as they pursue cloud migration.

Dependence on foreign platforms creates three categories of risk.

First, "kill switch" risk. Platform access can be cut off in the event of diplomatic conflicts or sanctions. When Russia invaded Ukraine in 2022, Western companies immediately suspended services to Russia. Microsoft, Apple, Google, and Cisco all withdrew from the Russian market. Palantir officially sided with Ukraine and terminated all cooperation with Russia. If Korea enters conflict with a particular bloc, what happens when core infrastructure depends on foreign platforms?

Second, technology lock-in risk. Once data is uploaded and systems built on a specific platform, migration to another is difficult. Transition costs are enormous, compatibility issues arise, and trained AI models are not easily transferred. This is why Palantir's Apollo platform exerts a powerful lock-in effect on customers. Once you enter the Palantir ecosystem, it is hard to leave.

Third, weakened bargaining power. Without alternatives, there is no room to negotiate prices or terms. When Britain's National Health Service (NHS) signed a 330-million-pound contract with Palantir in 2023, fierce opposition erupted domestically. The debate was whether it was appropriate to entrust NHS data to an American company. But the NHS was already in a state where it could not operate without Palantir's system. In a negotiation without alternatives, you cannot avoid the subordinate position.

Recognizing these risks, the Korean government has raised the discourse of "sovereign AI."

The National AI Strategy Committee has presented three principles: sovereignty over national goal-setting, sovereignty over infrastructure operation, and sovereignty over talent acquisition. The core is that Korea must possess the key capabilities needed for the AI era indigenously. The government announced plans to secure 10,000 GPUs by 2025, introduce a top-six-ranked supercomputer by the first half of 2026, and build a National AI Computing Center by 2027.

Infrastructure alone is insufficient. True sovereignty comes from the software capabilities running on that infrastructure.

Palantir's value does not reside in servers or GPUs. It lies in ontology technology that connects data and extracts meaning, the engineering capability to implement it, and the Forward Deployed Engineer (FTE) workforce that understands clients' problems and designs solutions. Even if Korea can make world-class semiconductors, if the software platform running on them comes from abroad, sovereignty remains incomplete.

The AI Basic Act passed in December 2024 is significant in this context. Its official title is the "Basic Act on Advancement of Artificial Intelligence and Establishment of Trust." Of 264 members present, 260 voted in favor. This law takes effect in January 2026. It contains transparency obligations for high-impact AI, safety assurance measures, and requirements for appointing domestic representatives. Foreign big tech companies must comply with domestic regulations to provide AI services in Korea.

But regulations alone cannot win the competition.

The ultimate answer is to build a Korean Palantir, a Korean data platform. A sovereign platform capable of securely integrating and analyzing government data, defense data, and public data. Without it, Korea will forever remain a consumer of foreign platforms.

(2) The Fusion Potential of K-Defense and AI Technology

The exhibit that drew the most attention at the ADEX 2025 show floor in October 2025 was LIG Nex1's electronic attack aircraft (EA aircraft) model.

The researcher giving the presentation told visitors: "This aircraft detects and analyzes radio signals across North Korea to jam communications or protect friendly operations. Domestic technology has reached the level of operations at 40,000 feet." On the screen, a demonstration showed electronic jamming signals spreading across a virtual battlefield map.

The evolution of K-defense is remarkable.

The negotiation volume for orders at ADEX grew from $29.4 billion in 2023 to $44.9 billion in 2025, a 52.7% increase. Hanwha Aerospace exported K2 tanks to Poland and K9 self-propelled howitzers to Norway. Seventy-two percent of Hyundai Rotem's defense solutions revenue comes from overseas. The government has set a target of $20 billion in defense exports by 2030.

But hardware alone cannot sustain a competitive edge.

The fastest-growing segment of the global defense market is software. It is symbolic that Palantir's market capitalization surpassed that of traditional defense giant Northrop Grumman. The U.S. Department of Defense's $10 billion investment in Palantir in 2025 was not for aircraft carriers or stealth fighters but for a data integration platform. The era has arrived where it is not the weapons but the systems connecting them that determine the outcome of war.

For K-defense to leap to the next level, AI software capabilities must be fused with weapons systems.

That movement has already begun. Hanwha Aerospace's "Chunmoo 3.0" is a system where reconnaissance drones and loitering munitions share intelligence in real time through AI collaboration. When the launcher opens, drones take off first to search for targets, and based on the acquired information, loitering munitions execute precision strikes. A Hanwha representative explained: "Each round collaborates in real time."

Hanwha Ocean's "Smart Battleship" uses AI to integrate the Combat Management System (CMS), Engine Control System (ECS), and Integrated Bridge System (IBS). Automatic target recognition is built in. A ship that typically requires 60 crew members can be operated with mid-30s personnel. A critical solution for the demographic cliff era.

LIG Nex1 developed an AI-based Battle Management System (BMS 2.0). It connects all weapons, sensors, and command systems on the battlefield into a single network to support real-time decision-making. Long-range air-to-surface and air-to-air guided missiles for the KF-21 fighter were also unveiled. The company has partnered with KAIST to begin developing AESA radar and SAR semiconductors.

Korea Aerospace Industries (KAI) has set a goal of completing a Manned-Unmanned Teaming (MUM-T) system by 2030. An AI pilot simulator and the development of the KUS-FS, which will serve as an unmanned wingman for the KF-21, are underway. It has also partnered with Samsung Electronics for AI semiconductor development.

What is noteworthy is that these technologies are being connected to export products.

Hanwha Systems signed a contract to supply Boeing with the ELAD (Electronic Laser Attack Demonstrator) for F-15 fighters. This is the first case of a Korean defense company entering the U.S. defense market. LIG Nex1's guided rocket "Bigong" passed the U.S. Foreign Comparative Testing (FCT), earning eligibility for U.S. export. KAI is pursuing the U.S. Navy trainer aircraft program, a large-scale contract worth some 10 trillion won.

But the true game changer is not individual weapons but system integration capability.

What Palantir demonstrated in Ukraine was not the performance of individual weapons. It was the shortening of the kill chain by connecting satellites, drones, ground forces, artillery, and missile units into a single data network. There is a world of difference between Korea exporting K2 tanks, K9 self-propelled howitzers, Chunmoo multiple rocket launchers, and Cheongung missiles individually, and exporting them together with an integrated platform connecting them all.

The "K-Air Defense Belt" vision presented by LIG Nex1 moves in this direction. It is a plan to export the Korean Air and Missile Defense (KAMD) lineup, composed of L-SAM, Cheongung-II, Haegung, and Singung, as a package together with AI-based command-and-control systems. The concept is to expand customized multi-layer air defense solutions to the Middle East, Asia, North Africa, and Europe, with technology transfer and local production models.

This is the potential for a Korean Palantir.

In hardware, Korea has already reached world-class levels. The remaining tasks are the software connecting that hardware, the platform integrating the data, and the talent to build it. It is a good sign that the four major K-defense companies saw operating profits in Q3 2025 surge 133.8% year-over-year to 3.5 trillion won. If these profits are reinvested in AI and software capabilities, K-defense can leap from a hardware exporter to a battlefield systems exporter.

C. Implications for Education, Administration, and Public Policy

(1) Public Data Governance and Applying the Palantir Model

In the spring of 2020, Britain's National Health Service (NHS) was in chaos.

COVID-19 patients were surging, but the NHS could not even determine how many beds were available and where. Thousands of healthcare institutions operated their own systems, and those systems did not talk to each other. Patient data, equipment inventories, staffing allocations, and vaccine supplies were scattered across separate databases. It was an information silo.

The NHS called in Palantir.

After the Foundry platform was introduced, things changed. Oxygen supply volumes, mask inventories, ventilator locations, and bed occupancy rates were displayed on a single dashboard. When vaccines arrived, AI calculated how many should be allocated to which region. The appointment system matched supply and demand in real time. An NHS official assessed: "For the first time, we could see the entire system at a glance."

What can Korea's public sector learn from this example?

The Korean government possesses one of the largest public data repositories in the world. Resident registration, health insurance, tax, real estate, vehicle, financial, and education data are meticulously accumulated. The problem is that these datasets are not connected. Each ministry operates its own system, and those systems are not interoperable. The databases of the Ministry of the Interior, the Ministry of Health and Welfare, the Ministry of Land, and the Ministry of Education each exist independently.

Under this structure, it is difficult to measure the effectiveness of policies. How long do unemployment benefit recipients stay employed after finding a job? How do the economic circumstances of families moving into public housing change? What impact does educational investment have on future income? Answering these questions requires linking data from multiple ministries. But under the current system, such linkages are extremely limited.

The core of the Palantir model is a "Single Source of Truth."

It integrates scattered data into a unified ontology so that the whole picture can be seen. Just as Hospital A's patient data was connected to Hospital B's equipment data in the NHS case, if Korea's welfare data were linked with employment data, policy blind spots could be identified and resources allocated more efficiently.

The National AI Strategy Committee, launched in September 2024, presented "nationwide AI transformation" as a core policy pillar. Applying AI to the public sector is an important agenda. AI-native government platforms, integrated citizen service platforms, and redesigning public systems using private-sector capabilities are under discussion. The Lee Jae-myung government has announced plans to build high-quality data cluster centers for AI learning and expand the opening of public data.

But data integration requires preconditions. Specifically, governance.

Who can access the data, for what purpose, how long it can be retained, and who is accountable. Without clear rules, data integration only creates chaos. This is why Palantir emphasizes "metadata and lineage systems." All data must be tracked , where it came from, who accessed it when, and how it was transformed , so that accountability is clear.

The "Guide to Personal Information Processing for Generative AI Development and Use" released by the Personal Information Protection Commission in August 2025 is important in this context. This guide divides the AI lifecycle into four stages , purpose setting, strategy formulation, learning/development, and application/management , and presents considerations for personal information protection at each stage. The core is establishing a governance framework centered on the CPO (Chief Privacy Officer).

To apply the Palantir model in the public sector, it is necessary to go beyond simply adopting a platform and redesign the governance framework. Inter-ministry data sharing principles, access rights management systems, audit trail mechanisms, and accountability systems for misuse , these elements must be in place to reap the benefits of data integration while controlling the risks.

(2) Finding the Balance Between Efficiency and Privacy Protection

In February 2023, the German Federal Constitutional Court delivered a historic ruling.

The Palantir-based systems used by police in Hesse and Hamburg were unconstitutional, the court decided. The court declared: "This system enables profiling of individuals with a single click. This violates the right to informational self-determination." Hesse was required to amend its law by September 2023, and Hamburg's relevant law was invalidated.

This ruling reveals the dual nature of Palantir technology.

In terms of efficiency, Palantir is overwhelming. In criminal investigations, it can connect hundreds of databases to track suspects. In counter-terrorism, it can detect anomalies in advance. In public health crises, it can optimize resource allocation. All of this is possible because it allows you to "see the whole picture."

But that very capability becomes a threat to privacy.

Seeing the whole picture also means scrutinizing every detail of an individual's life. The LAPD's predictive policing program "Operation LASER" predicted the locations and people where crime would occur and deployed police accordingly. It was effective. But controversy erupted when the algorithm concentrated police in neighborhoods of particular racial groups, reproducing racial discrimination. Prediction reproduced bias.

Korea is not free from this dilemma.

Korea is one of the countries with the most tightly woven personal information infrastructure in the world. Resident registration numbers track individuals from the moment of birth for a lifetime. Health insurance data contains all medical records. Tax data details income and assets. Communications data reveals location and behavioral patterns. What becomes possible if AI connects all of this?

The AI Basic Act passed in December 2024 foregrounded "trust."

This is why the law's official title is the "Basic Act on Advancement of Artificial Intelligence and Establishment of Trust." The law presents three fundamental principles: the principle of human dignity, the principle of public good, and the principle of technological fitness for purpose. It also specifies ten core requirements: human rights protection, privacy protection, respect for diversity, prohibition of infringement, public interest, solidarity, data management, accountability, safety, and transparency.

Special obligations are imposed on "high-impact AI." High-impact AI is AI that can have significant effects on human life, safety, or fundamental rights. Autonomous driving systems, medical diagnostic AI, and financial credit scoring AI fall into this category. Operators of such AI bear transparency obligations, safety assurance obligations, and AI impact assessment obligations. Foreign big tech companies providing services in Korea must designate a domestic representative.

But law alone is not enough to achieve balance.

Technical mechanisms are needed. The "Human-Machine Teaming" concept emphasized by Palantir offers useful reference. AI must not operate as a black box; humans must be able to understand and review why a particular decision was made. Explainable AI (XAI), audit trail systems, and bias detection tools are essential.

Institutional mechanisms are also necessary. The role of the Personal Information Protection Commission must be strengthened. It must be able to monitor what data was used and how when AI is employed in public decision-making. Procedures must be established for citizens affected by algorithmic decisions to file objections. Just as the EU AI Act specifies the right to "human oversight," Korea must build similar institutions.

Above all, social consensus is needed.

Where to set the trade-off between efficiency and privacy. How much surveillance to permit for crime prevention. How much personal data linkage to allow for improving public services. There are no correct answers to these questions. Every society will arrive at different answers.

Germany is sensitive to surveillance because of historical experience. It remembers how the Nazi Gestapo and the East German Stasi abused personal information. That is why the Constitutional Court put the brakes on the Palantir system. The United States chose to prioritize security after 9/11. That is why Palantir became a core tool of intelligence agencies.

What choice will Korea make?

The North Korean threat, the demographic cliff, and the demand for administrative efficiency strengthen the temptation of data integration. But memories of authoritarian history, experiences with personal information breaches, and growing citizen awareness of privacy push in the opposite direction. The AI Basic Act seeks this balance with the expression "trust-based." But how the principles written in the law will be implemented in reality remains unknown.

Alex Karp of Palantir has said: "Our technology sometimes kills people." He did not hide this fact. He acknowledged both the power and danger of technology simultaneously. What Korea should learn from the Palantir model is not just technology. It is honest dialogue about the impact that technology has on society, awareness of the need for democratic control, and the wisdom to manage the ceaseless tension between efficiency and freedom.

Harnessing the power of data without being consumed by it. That is Korea's true challenge on the path to the AI G3 leap.