A narrative chronicle of missions, victories and public dilemmas

Aboard a C-130 Hercules flying over Helmand Province, an infantry captain hurriedly leafs through some printed reports.

Every page carries a QR code: when he scans it, his tablet opens a map where the dusty roads blink red, yellow or green depending on the likelihood of encountering improvised explosive devices.

Whoever color-codes those routes does not wear a uniform; he works thousands of kilometers away, in a nondescript office in Denver, connected to the same database that feeds the troops. Behind the screen is a platform called Gotham and, behind Gotham, the open question of how much power a state should delegate to the algorithmic scoring of war.

Afghanistan, Mosul and Los Angeles: the first deployments

The baptism of fire: Afghanistan:

Palantir's first contracts with the U.S. Department of Defense seemed modest: pilot licenses so forensic analysts could look for patterns in IED detonations.

In 2010, however, the commander of the 82nd Airborne Regiment asked for Gotham to be deployed “forward,” inside Bagram base. The result was a system that, every early morning, ingested SIGACT messages, patrol logs and intercepted calls to predict where explosions would occur at dawn.

The predictions were not infallible oracles, but they forced a rethink of routes and schedules. According to an internal Army TRADOC report, units that used Palantir cut their IED casualties by 25% over nine consecutive months. The platform thus became part of the standard kit in Afghanistan and later in Iraq: as ubiquitous as body armor, as intangible as a radio rumor.

Urban operations: from the alleys of Mosul to the neighborhoods of Los Angeles:

In 2016 the battle for Mosul tested the software's elasticity. Soldiers of the Special Operations Joint Task Force tagged drone videos with street names; minutes later, another operator, 8,000 kilometers to the west, could overlay the clips on Ottoman-Turkish sewer plans. Gotham worked as a “hybrid brain” that absorbed the chaos of the old city and gave it back in the form of advance routes with lower collateral risk.

That same year, an ocean away, the Los Angeles Police Department began using the same technology—now dressed in the discreet name “LASER”—to identify hotspots of armed violence. The algorithm built “hot boxes” of 150 × 150 meters and assigned a risk score to individuals who, with no serious record, appeared as possible spreaders of crime. The project was canceled in 2020 after criticism from the ACLU, but it left a question buzzing in the offices: when can a weapon of war be called a tool of community prevention?

Interagency intelligence and Britain's NHS

Interagency intelligence: the Thursday oval table.

Across the Potomac, every Thursday at 7 a.m. a small group of analysts meets in a carpeted, phone-free room. There, on a 90-inch touchscreen, the same graph parades by that the Marines used in Helmand and homicide detectives used in Chicago.

The layers overlap: port logistics, financial flows, conversations intercepted under court order. Under an undersecretary's thumb, the nodes change color to the rhythm of the topics: drug trafficking, cyberattacks, pandemics.

The legend reads “Powered by Palantir.” Yet most of the people who make decisions do not know the code that prioritizes one lead over another. They trust the briefings—and the “audited” seal—to convince the President that an oil tanker deserves aerial surveillance or that a ransomware signal actually comes from a teenager in his bedroom.

NHS COVID-19 Data Store: healthcare as a national security exercise.

April 2020: the United Kingdom is in agony in the first wave of COVID-19. Hospitals at the edge of their capacity; ventilator stocks changing hour by hour. The National Health Service signs an emergency contract with Palantir to build, within days, the NHS Data Store.


The internal account describes a logistical war: ambulances sharing real-time telemetry, factories reporting PPE inventory, morgues updating occupancy rates. Foundry—not Gotham, this time—integrates everything into a dashboard where each red dot is a free ICU bed or an imminent bottleneck. Three weeks later, the director of health operations declares that the system “probably saved thousands of lives.” Privacy advocates reply that a lifesaver does not justify storing medical records in private hands.

The contract, renewed in 2022, continues to fuel headlines and demands for transparency.

ICE and Ukraine: two uses under scrutiny

Border surveillance and the ghost of ICE:

In the United States, Palantir maintains two lower-profile products: FALCON-SA and FALCON-TIP. Both allow Immigration and Customs Enforcement (ICE) to link detention databases, driver's license records and social media to locate people suspected of violating immigration laws.

A New York Times report revealed that, between 2017 and 2019, these systems were used to plan mass raids. Immigrant rights advocates argue that the tool turns administrative violations into priority detention targets, confusing terrorism with mere irregularity. Palantir responds that it only provides software and that end use is a matter of public policy. The debate reaches Congress: is it lawful for an algorithm to determine the urgency of separating a father from his children?

Ukraine: war in high resolution.

In February 2022, as Russian troops crossed the border, a team of Palantir engineers installed classified Gotham nodes for the Ukrainian General Staff. The platform absorbed commercial satellite imagery, Telegram posts and hacked railway manifests.

What was unprecedented was the speed of fusion: in under 30 minutes, a drone video entered the system, was georeferenced and appeared as “linked evidence” on a convoy's route. Ukrainian officers say that cycle cut artillery response time from hours to minutes. Moscow accuses the West of “outsourcing” its intelligence to a company. Palantir stays silent, aware that its brand already appears in scientific publications on “algorithmic warfare.”

European regulation, dependence and ethics

Transparency vs. opacity: the regulatory pendulum.

In Brussels, the draft Artificial Intelligence Regulation includes a section on “general-purpose systems used by law enforcement.”

Several NGOs are pressing for Palantir to fall into the “high-risk” category, subject to code audits and export limits.

The company counters: opening the engine up completely would expose tactics to adversaries. Between the black box and total glass, the EU is looking for a middle point where states can trust and citizens need not fear.

The hearings are an unexpected theater: programmers explain hashes and graphs to lawyers who barely command the computing jargon. Even so, the final decision—publish or conceal—will determine whether Gotham and Foundry keep expanding in Europe or run into a regulatory wall as firm as the medieval fortresses that still dot the continent.

Defense or dependence?

With each new contract, the company reinforces its role as “sovereign data infrastructure”: a seductive concept for governments that lack technological muscle.

But in the corridors of the Pentagon and NATO a recurring worry arises: what happens if Palantir raises prices, changes hands or decides not to renew licenses in the middle of a crisis? Strategic dependence—the famous lock-in—is not only economic; it is tactical.

To mitigate the risk, some states develop in-house clones; others demand code deposits in escrow. Meanwhile, Palantir signs multiyear agreements that include training and “offline” tools to calm fears. The balance holds, fragile, thanks to tangible benefits: fewer casualties, faster logistics, evidence-based decisions.

Under the lens of ethics:

In 2023 an academic consortium analyzed 2,400 court documents and concluded that reports generated by Gotham are cited with increasing frequency as evidence in military courts. Does that mean an algorithm influences the attribution of guilt? Palantir maintains that its software is “a magnifying glass, not a judge.” Yet the line blurs when a prosecutor shows the jury a multicolored graph that seems to point to guilt in fluorescent strokes.

The company has created a “Responsibility Council” that includes philosophers and former judges. Its opinions are advisory, never binding. Critics compare them to an in-house tobacco review board weighing in on nicotine.

The essential question remains: can a private corporation arbitrate its own impact on civil rights?

Epilogue: the Möbius strip between the front and home:

Let us return to the captain on the Hercules. Upon landing at Camp Bastion, he confirms that the suggested route avoided the IED that went off 300 meters away.

Hours later, in Los Angeles, a patrol officer reads a report that rates a young man as a “moderate-risk actor” because his cousin was arrested in Compton. And, that same night, a British doctor finds an ICU bed thanks to a dashboard that refreshes every 15 seconds.

Three stories, one common thread: Palantir turns scattered data into maps for action. The benefit is palpable; so are the risks.

From here on, the question is no longer whether the state will use the platform, but under what conditions it can keep using it without compromising the delicate pact between security and freedom. On that edge, in real time, is being written the next chapter of a company that began by tracking terrorists and ended up managing pandemics.