PALANTIR: War, Surveillance, Artificial Intelligence

Part 5: Controversy and Regulation, The Shadow of the Surveillance State

Chapter 9. Big Brother's Eye: Surveillance and Predictive Policing

Attorney Kyungjin Kim

A. The Reality of Minority Report

(1) The Achievements and Controversies of LAPD's Predictive Policing Program (Operation Laser)

One day in 2011, a police officer in the Newton Division of the Los Angeles Police Department (LAPD) was staring at a tablet inside his patrol car. On the screen was a photograph of a young Black man. Name, address, tattoo locations, vehicle license plate number, and a number written beside them. 15 points. This young man had been classified as a "Chronic Offender." The officer's mission was to find this person and make contact. What crime he had committed did not matter. The algorithm had designated him as a dangerous individual.

This was how Operation LASER worked. LASER stood for "Los Angeles Strategic Extraction and Restoration." Translated literally, it means "Los Angeles Strategic Extraction and Restoration." The people who designed the program embedded a medical metaphor in the name. The official documents from the Bureau of Justice Assistance under the federal Department of Justice stated the following: "Operation LASER can be likened to laser surgery. Just as a trained physician uses modern technology to remove a tumor or improve eyesight."

Tumor. They called certain citizens tumors.

Palantir's Gotham platform served as the brain of this program. The system integrated dozens of data sources, including arrest records, field interview reports, automatic license plate reader data, gang databases, and social media activity. And it assigned a score to each individual. Five points if they had previously possessed a firearm. One point if they were on probation. One point was added each time they had contact with police. Once a score exceeded a certain threshold, the person's name was placed on the "Chronic Offender Bulletin." Like a wanted poster, photographs and personal information were distributed to patrol cars.

The problem lay in the operating logic of this system. If the score went up with every police contact, then people living in areas where police frequently patrolled automatically became high-risk individuals. And where did police patrol frequently? Areas where many crimes had been reported in the past. This created a self-fulfilling prophecy loop. Police went more often to poor Black and Hispanic neighborhoods, residents of those areas were stopped and searched more frequently, so they received higher scores, and so they became surveillance targets more often.

In March 2019, the audit results from the LAPD Inspector General's office were released. The results were shocking. Of the 233 people classified as chronic offenders, 84 percent were Latino or Black. These two groups accounted for less than 60 percent of Los Angeles's total population. There were even more surprising facts. Nearly half of the chronic offenders had never been arrested for a violent crime, or had been arrested only once. And about 20 percent, one in five, had no record whatsoever of "meaningful contact" with police.

The audit report pointed to another problem. It was impossible to demonstrate the program's effectiveness. Of the 13 areas designated as LASER zones, six areas actually saw increases in violent crime or showed no difference from non-LASER zones. Data collection methods were inconsistent. Some officers diligently entered their contact records, while others did not. There was no way to verify the accuracy of the system itself.

Hamid Khan, co-founder of the Stop LAPD Spying Coalition, had fought against this program for years. His organization secured internal documents through public records request lawsuits and organized protests with community residents. Khan said in an interview with MIT Technology Review: "We officially launched the campaign in 2016. The goal was to understand the program, inform the community, and dismantle it."

In April 2019, LAPD finally officially terminated Operation LASER. But the story did not end there.

(2) Algorithmic Bias and Racial Discrimination Issues

The very month Operation LASER was abolished, LAPD announced a new program. It was called "Data-Informed Community-Focused Policing" (DICFP). Only the name had changed.

In 2021, new documents obtained by the Stop LAPD Spying Coalition were made public. The documents showed how LASER and another predictive policing program called PredPol had automated and reinforced existing racial biases. PredPol was a location-based system that predicted 500-foot by 500-foot areas with a high probability of crime based on past crime data. It was an algorithm inspired by earthquake aftershock prediction models.

Sarah Brayne, a sociologist at the University of Texas at Austin, participated as an embedded researcher at LAPD for several months and observed how predictive policing actually worked. Her book, Predict and Surveil, contains shocking scenes. One sergeant explained to her: "There are a lot of minor violations to stop someone for. Yesterday, this person may have been stopped for jaywalking."

Brayne also witnessed how police used automatic license plate readers installed in front of emergency rooms. Family members or friends of shooting victims often drove the patient to the hospital and left quickly. Police tracked those license plates to build the victim's social network. Even without evidence of a crime, people who entered the network became surveillance targets.

Andrew Ferguson, a law professor at American University and author of The Rise of Big Data Policing, defined the essence of the Palantir system this way: "Palantir is less a data-driven policing system than a data-driven surveillance system. Palantir helps police surveil people, not keep people safe."

In 2023, the nonprofit investigative outlet The Markup analyzed the actual performance of PredPol (later rebranded as Geolitica). After reviewing more than 23,000 predictions over the course of 2018, the success rate was less than 0.5 percent. Fewer than 100 cases saw a predicted category of crime actually occur at the predicted location and time and be reported to police. The Plainfield police chief told The Markup that the money spent on Geolitica software "could have been better spent elsewhere."

In April 2020, LAPD announced that it was terminating its PredPol contract, citing budget constraints due to COVID-19. But critics pointed out that this was not the real reason. Years of accumulated evidence showing the program was ineffective and discriminatory had led to this result. The Palo Alto Police Department had already terminated its contract after three years of using PredPol, saying it "didn't get any value."

The core of the problem is not that algorithms create bias, but that they amplify existing bias. Shakeer Rahman, an activist with the Stop LAPD Spying Coalition, explained it this way: "Data-driven policing automates existing police logic. That includes targeting poor people, targeting unhoused people, targeting Black and brown and disabled people. Now this is helping automate those practices, automate harm, automate deportation, automate the exclusion that police have always been in the business of."

In February 2024, the city of Chicago refused to renew its contract with ShotSpotter, a gunshot detection and predictive policing system. Mayor Brandon Johnson stated in a press release that "the City of Chicago will deploy resources toward strategies and tactics proven to be effective in accelerating the current trend of declining violent crime." Skepticism toward predictive policing was spreading.

Yet Palantir was still thriving. Through COVID-19 tracking and vaccine safety analysis contracts, it had gained entry to the National Institutes of Health (NIH) and the Food and Drug Administration (FDA). Since its stock market listing in September 2020, its share price had more than tripled. The failure in Los Angeles seemed to have no impact on the company's growth.

B. ICE and the Deportation Machine

(1) Supporting Immigration and Customs Enforcement's Tracking of Undocumented Immigrants

One day in April 2025, an agent in the Enforcement and Removal Operations division of U.S. Immigration and Customs Enforcement (ICE) logged into a new system. On the screen appeared the name "ImmigrationOS." This platform, developed by Palantir, gave the agent remarkable capabilities. With a few clicks, they could view a specific individual's passport records, Social Security files, IRS tax data, driver's license information, and even automatic license plate reader data at a glance. The system could track voluntary departures "in near real-time," identify visa overstays, and designate individuals associated with transnational criminal organizations like MS-13 or Tren de Aragua as targets.

This was the deportation machine of the Trump second-term administration.

The relationship between Palantir and ICE dates back to 2014. At the time, Palantir built a tool called Falcon for ICE's Homeland Security Investigations (HSI). Falcon enabled agents to track specific individuals' locations over time by cell phone number, analyze air travel records, and share field interview information in real time. This contract, signed during the Obama administration, continued through the Biden administration.

But the ImmigrationOS of 2025 was on a different level. In mid-April 2025, ICE signed a $30 million contract with Palantir. According to federal contract records, a prototype was to be delivered by September 25, 2025. In the contract justification documents, ICE stated there was an "urgent and compelling need." And it declared Palantir was the "sole source."

"Palantir possesses deep institutional knowledge of ICE operations," the documents stated. "The company is already collecting and processing data from multiple federal agencies, placing it in a unique position to develop a prototype within six months." It warned that any delay would undermine the enforcement mission specified in President Trump's executive orders.

ImmigrationOS was designed to provide three core functions. First, target selection and enforcement prioritization. This function helps determine who to deport first, including "violent criminals," gang members, and visa overstays. What data or criteria are used to identify "violent criminals" was not disclosed. Second, voluntary departure tracking. It monitors whether individuals leave the United States voluntarily. Third, immigration lifecycle management. It simplifys the entire process from identification to deportation, minimizing time and resource waste.

In September 2025, ICE signed an additional contract worth $29.9 million. It was for software license renewal, operations and maintenance, and adaptive maintenance support. The contract was valid through September 2027. According to public records obtained by the New York Times, the total value of federal contracts awarded to Palantir since Trump's inauguration exceeded $900 million. This was nearly double the $541.2 million from 2024.

The system's power lay in its data integration capability. ImmigrationOS pulled in not only government databases but also information from the private sector. Credit information, utility records, Department of Motor Vehicles data, and airline passenger information were all connected. According to analysis by the Migration Policy Institute, ICE is believed to be among the largest government purchasers of commercial credit, utility, and vehicle registration information.

In April 2025, Palantir's Chief Technology Officer (CTO) and President of Palantir USG, Akash Jain, wrote in an internal Slack message: "A quick update on our work with ICE." According to the leaked message, the company had developed "a new set of data integrations and workflows with ICE" in response to "the new administration's focus on using data to drive enforcement operations."

According to CNN reporting, ImmigrationOS would use artificial intelligence systems to allow immigration agents to "authorize raids, book arrests, generate legal documents, and route transfers to deportation flights or detention facilities from a single interface."

(2) Internal Employees' Ethical Backlash and Exodus of Resignations

In May 2025, thirteen former Palantir employees published an open letter. The title was "The Scouring of the Shire." It referenced the scene in The Lord of the Rings where the hobbits' homeland is devastated even after Sauron's defeat. Considering that the company's name derived from Tolkien's world, there could be no sharper criticism.

"Companies are kowtowing to the Trump administration, suppressing dissent, and aligning with his xenophobic, sexist, plutocratic agenda," the letter stated. They warned that the company's ethical guardrails, originally designed to prevent discrimination, disinformation, and abuse of power, were being dismantled.

This backlash was not new. CEO Alex Karp frankly acknowledged it in a March 2024 CNBC interview. "We have lost employees. We will lose more." He continued: "If you have a position that doesn't cost you employees, it's not a position."

Brianna Katherine Martin was one of the employees who left the company over ethical concerns. She said projects like ImmigrationOS constituted "complicity in human rights and constitutional violations."

Criticism from human rights organizations was even fiercer. Mariana Olaizola Rosenblat, a policy advisor at NYU Stern's Center for Business and Human Rights, said in an interview with Axios: "By supporting the Trump administration's deportation apparatus, Palantir is complicit in those human rights and constitutional violations."

The American Immigration Council pointed out fundamental problems with the system. ImmigrationOS pulls information from government databases regardless of their accuracy or reliability. Outdated or incorrect information can lead to life-altering consequences. Detention, loss of status, or wrongful deportation.

One of the most uncomfortable facts was the conflict of interest issue. Stephen Miller, the chief architect of the Trump administration's immigration policy, held significant financial stakes in Palantir. The person designing the government's immigration policy was a shareholder in the company providing the technology to enforce that policy. Was this a coincidence, or a designed outcome?

In December 2025, the Washington Post reported on how Palantir had pivoted. The article's lead read: "For years, Alex Karp declared that his data management firm was 'engaged in supporting progressive values' and 'repeatedly walked away' from contracts it deemed targeting minorities or unethical."

But now the company had become the technological backbone of the most aggressive mass deportation campaign in American history. Palantir argues that it merely provides tools. It is not the company that decides who to target, deport, and surveil, they say.

But critics push back. Design choices shape real-world outcomes. AI architecture becomes policy. When an algorithm classifies certain groups as higher risk, that is not a technical decision but a political one.

Likhita Banerji, deputy director of Amnesty Tech at Amnesty International, told El País: "These technologies can systematically promote racism, discrimination, and oppression, and are routinely used to advance racist and xenophobic agendas."

C. Lavender and the Gaza Strip

(1) Official Declaration of Support in the Israel-Hamas War

One day in January 2024, an unusual gathering took place in a conference room in Tel Aviv. It was the first time the Palantir board of directors held a meeting outside the United States, and in Israel no less. Peter Thiel and Alex Karp sat alongside Israeli Ministry of Defense officials for a photograph. Danny Gold, director of the Defense Research and Development Directorate (DDR&D), known for leading the development of the Iron Dome missile defense system, was also present.

The photograph did not appear in the Israeli press. But Bloomberg reported it. The news was that Israel had agreed to purchase an AI-based system called AIP from Palantir. This system could support decision-making based on intelligence, analyze enemy targets, and suggest combat actions. The contract was expected to be worth tens of millions of dollars.

Alex Karp did not hide his position. In a February 2024 earnings conference call, he said: "I am incredibly proud that within weeks of October 7th, we were on the ground and involved in operationally significant activities in Israel."

Karp became even more candid during a conversation with legendary Israeli entrepreneur Yossi Vardi at Tel Aviv University. "There is nobody who doesn't have a position on Israel, if you're a big company. There are plenty of people who are not as pro-Israel as I am, but they think of Israel as a very special place and they are more sympathetic to Israel's position and appreciate what Israel has accomplished in building a state in the desert."

The company took out a full-page advertisement in the New York Times. "Palantir stands with Israel."

The relationship between Palantir and the Israeli military establishment had begun much earlier. In 2015, the company opened an office overlooking Rothschild Boulevard and Yehuda Halevi Street in Tel Aviv. This was a strategic move to deeply integrate into the Israeli tech ecosystem. Karp himself acknowledged it in a December 2023 interview with Fox Business: "We are very well known in Israel. Israel thinks highly of our products."

But the timing of the January 2024 agreement, which Bloomberg described as "a strategic partnership for battlefield technology," raised questions. This agreement was concluded at a point when the carnage in Gaza had already been underway for three months. The Business and Human Rights Resource Centre asked: "Was it a calculated move by Palantir to use an escalating conflict as an opportunity to test its AI models against civilians? Did it turn Gaza into a horrific proving ground for technology?"

(2) AI Target Identification Systems and the "War Crimes by Algorithm" Debate

In April 2024, the Israel-Palestine specialist outlet +972 Magazine and the Hebrew-language outlet Local Call published a shocking investigative report. The headline read: "'Lavender': The AI Machine Directing Israel's Bombing Spree in Gaza." Six Israeli military intelligence officers testified on condition of anonymity.

Lavender was an AI-based database designed to mark all suspected members of Hamas and Palestinian Islamic Jihad (PIJ) military organizations, including low-ranking operatives, as potential bombing targets. The system analyzed virtually everyone in the Gaza Strip and assigned a score from 1 to 100. That score indicated the likelihood that the person was actively involved with an armed group.

According to the sources, during the first few weeks of the war, the Israeli military relied almost entirely on Lavender. The system marked 37,000 Palestinians as suspected combatants, and their homes as potential airstrike targets.

Even more shocking was the human review process. The intelligence officers who testified said they spent an average of 20 seconds approving a target generated by Lavender. In many cases, the verification amounted to checking whether the target was male. Under the assumption that women were not combatants.

Everyone knew the system had an error rate of about 10 percent. But it was considered an acceptable level. One source told +972: "We were not interested in killing Hamas operatives only when they were in a military building or engaged in military activity. On the contrary, the IDF bombed them in homes without hesitation, as a first option. It's much easier to bomb a family's home. The system is built to look for them in those situations."

There was another system linked to Lavender. It was called "Where's Daddy?" This system tracked individuals marked by Lavender until they entered their homes. It deliberately targeted nighttime, when families would be together.

The acceptable civilian casualty thresholds were also revealed. For killing a single low-ranking Hamas operative, the deaths of 15 to 20 civilians were permitted. For senior commanders, that number rose to between 100 and 300.

One source testified: "The targets are endless." Once the bombing started, commanders knew "there are another 36,000 waiting."

In May 2025, evidence recorded in classified Israeli military databases was made public. Of the 53,000 Palestinians killed in Gaza, only 17 percent were combatants. This means 83 percent were civilians. The UN verified at the end of 2024 that approximately 70 percent of the dead were women and children.

UN Secretary-General António Guterres said he was "deeply concerned." He pointed out that Israel's use of artificial intelligence in military operations endangered civilians and obscured accountability.

Ben Saul, the UN Special Rapporteur on human rights and counter-terrorism, was more direct: "If the reports about Israel's use of AI are true, many Israeli airstrikes in Gaza would constitute war crimes of disproportionate attacks."

Marc Owen Jones, a professor at Hamad bin Khalifa University, said of the Lavender system: "Let's be clear. This is AI-assisted genocide, and there should be calls for a moratorium on the use of AI in warfare going forward."

Whether Palantir developed Lavender has not been confirmed. It is reported to have been developed by Unit 8200, the Israeli military's intelligence unit. But the timing of the January 2024 strategic partnership, and the fact that Palantir's technological capabilities could be applied to similar systems, raises questions.

What is certain is that Palantir actively supported this war. In a statement submitted to the Business and Human Rights Resource Centre, Palantir acknowledged providing "expanded, critical target identification capabilities" to the Israeli military. Those capabilities were likely used to carry out devastating strikes in Gaza, and possibly beyond.

Karp appeared visibly shaken when questioned about Lavender at a Cambridge University debate in May 2024. Investor Catherine Austin Fitts observed: "Peter Thiel may be legally protected, but I think he's beginning to realize that a reckoning could come in the court of public opinion."

But Thiel's position remained unshaken. "I broadly believe that the IDF gets to decide what it wants, and that they are broadly right."

Protests continued in front of Palantir headquarters, at the Hill and Valley Forum in Washington, D.C., and on university campuses worldwide. Palestinian-American Sumer Mobarak rushed the stage during a Karp speech, shouting: "You are making money off of killing Palestinian people. Palantir kills Palestinian people with their AI and their technology. You are killing my family in Palestine."

Karp's response was this: "The primary reason people are dying in Palestine is because Hamas figured out it has millions of 'useful idiots.'"

The video of this scene was viewed millions of times. But Palantir's stock price continued to climb. Market capitalization exceeded $200 billion. Amid the debate over war crimes by algorithm, investors were making money.

Who watches the algorithms. That is the question this era poses.