In Minority Report (2002), Tom Cruise plays an LAPD cop in a “pre-crime” unit that arrests people before they commit a crime. In the film, the predictions come from mutated “pre-cogs.” In real life, the forecasts come from data—lots of it.
Few companies symbolize that shift more than Palantir, a data-analytics firm founded in 2003 by Peter Thiel, Alex Karp, Stephen Cohen, Joe Lonsdale, and Nathan Gettings. Early on, it received backing from In-Q-Tel, the CIA’s venture-capital arm. Palantir’s work has long been tied to national security, but the tools honed for war zones and intelligence work have increasingly shown up in American cities—particularly in the push for “predictive policing.”
The question is whether those methods make communities safer—or whether they harden suspicion, amplify bias, and pour fuel on already tense relationships between police and the public.
From counterinsurgency to city streets
Palantir built a business around joining messy, fragmented data into something searchable and actionable. Some of its earliest and most influential use cases emerged from the post-9/11 national-security world, where agencies and the military needed faster ways to connect dots across people, places, communications, and events.
Reporting based on a leaked 2013 Palantir document described deployments in Iraq and Afghanistan where analysts used software to identify patterns—like links between roadside bomb incidents and potential triggering mechanisms—and to speed up identification workflows by connecting field data to existing biometric and intelligence records.
That kind of capability has obvious appeal: faster analysis, tighter coordination, fewer blind spots. It also invites a dangerous leap in language—from “finding patterns” to “predicting the future.”
The seductive promise of “predictive” policing
In policing, predictive systems generally come in two flavors:
Place-based prediction (hotspots)
These models use historical crime data—location, time, day of week—to forecast where crime is more likely to occur. The output is often a “hotspot” map that guides patrols.
Person-based targeting (risk lists)
Other programs attempt to identify individuals who might be at higher risk of becoming involved in violence—either as perpetrators or victims—based on networks, prior contacts, and other indicators.
Those are not the same thing, but they’re frequently blended together in the public imagination. And that’s where the Minority Report comparison becomes tempting—and misleading.
Palantir has been used by the Los Angeles Police Department in programs widely reported to include Operation LASER, which generated targeting lists and risk scores. At the same time, LAPD and other departments have used separate predictive policing tools (such as PredPol) that focus on forecasting locations. However you slice it, these systems share a common output: they nudge police attention toward certain neighborhoods and certain people—often repeatedly.
That nudge can become a loop.
When data becomes destiny
Predictive policing advocates argue that algorithms can help deploy limited resources, reduce response times, and prevent harm. Critics counter that these systems can encode existing biases into a veneer of mathematical objectivity.
If historically over-policed neighborhoods generate more stops, more searches, and more arrests, they also generate more data. Feed that data back into a model, and the model “learns” that these areas are high-risk—sending police back again. The cycle repeats.
For communities already living with heavy police presence, an algorithmic justification can feel like a new way to rationalize old assumptions. And when officers arrive primed by a system telling them this block, this person, this car is “high risk,” every encounter starts closer to the edge.
Militarized tools, domestic consequences
The deeper worry isn’t just that police departments are buying sophisticated software. It’s that the mindset and tactics developed for counterinsurgency—where the environment is treated as hostile territory and civilians can be potential threats—can seep into domestic law enforcement.
Ana Muniz, an activist and researcher with the Inglewood-based Youth Justice Coalition, has warned that as military and domestic policing grow more similar, the lines blur. The military’s mission is to defend territory from external enemies. Police are supposed to protect the public—including the people they’re policing.
When those roles start to resemble each other, trust collapses first—and safety can follow.
A company built for secrecy, now operating in public
Palantir has always projected an aura of guardedness—work tied to sensitive programs, deployments in secure environments, and tight control over how its tools are described. It’s also a company that, despite once signaling it might avoid going public, ultimately went public in 2020 via a direct listing—bringing more visibility to its finances and contracts than it had in the past.
Yet the core tension remains: systems designed to connect data at scale are extraordinarily powerful, and that power doesn’t come with a built-in moral compass. The same analytics that help investigators spot financial fraud or find trafficking networks can also be used to expand surveillance, widen suspicion, and justify aggressive policing—especially when departments and vendors market the work as “prediction.”
Will predictive policing calm conflict—or intensify it?
Supporters see predictive tools as modern policing: informed, strategic, data-driven. Critics see a feedback machine that deepens disparities, legitimizes over-policing, and creates more confrontations in the communities most likely to be harmed by them.
Even in Minority Report, the most unsettling idea isn’t that the system predicts crime. It’s that people stop questioning the system—and start acting as if the system can’t be wrong.
When policing decisions are shaped by black-box scores and hotspot maps, accountability becomes harder, transparency becomes rarer, and the public is asked to trust not just an officer’s judgment—but an algorithm’s.
And in places where trust is already thin, that’s not a neutral change. It’s an accelerant.