Products are selected by our editors, we may earn commission from links on this page.
A data company born from the fear of terrorism after September 11 is now facing something it never anticipated: its own employees asking whether it has become the threat it was built to stop. Palantir Technologies, the surveillance software giant co-founded by billionaire Peter Thiel with early CIA backing, has long sold itself as a guardian of civil liberties. That story is now crumbling from the inside.
For two decades, Palantir employees absorbed the awkward dinner-table conversations that came with working for a company named after J.R.R. Tolkien’s corrupting all-seeing orb. The company’s pitch was that it used powerful data tools in service of good: fighting terrorism, protecting the vulnerable, acting as a check on abusive government. That pitch gave workers a framework for the discomfort. But that framework is now gone.
Two former employees reconnected by phone last year. One opened with a question before pleasantries were exchanged: “Are you tracking Palantir’s descent into fascism?” The second employee recalled the greeting plainly, saying the feeling inside the company was not that the work had become difficult or unpopular, but that it had become wrong. And that distinction matters. Difficulty can be reasoned through. Wrong sits differently.
This article was created with the assistance of AI and reviewed by our editorial team for accuracy and clarity.
The Contract That Started It All
Palantir’s deepening role in immigration enforcement hardened into a formal arrangement in April 2025, when ICE signed a $30 million contract with the company to build a platform providing near real-time tracking of migrant movements across the United States. That contract has since expanded significantly, ballooning to over $145 million, with a prototype of the system, dubbed ImmigrationOS, scheduled for delivery to the agency. For workers already uneasy, the dollar figure and the name made it impossible to look away.
Palantir had also been secretly providing ICE with a tracking tool called “ELITE” designed to scour Medicaid data and other government sources to build profiles of individuals who could potentially be deported, generating a detailed dossier for each person that included their location. This was not the work of a company threading a careful ethical needle. It was the work of a company that had chosen a side, and employees knew it.
The breaking point came in January 2026 after federal agents shot and killed Alex Pretti, a nurse, during protests against ICE activity in Minneapolis. Workers flooded a company-wide Slack channel with demands for answers. “In my opinion ICE are the bad guys. I am not proud that the company I enjoy so much working for is part of this,” one worker wrote in the thread. The comment captured what many had been feeling but could not yet say aloud.
“A Sufficiently Malicious Customer Is Basically Impossible to Prevent”
Management’s response to the internal uproar was a series of ask-me-anything forums. One, arranged independently by two team leads, including one who had worked directly on the ICE contract, exposed just how limited Palantir’s guardrails actually were. During the session, a member of the company’s own privacy and civil liberties team acknowledged that if ICE chose to misuse the tools, Palantir could do little to stop it in real time. The admission landed hard.
The employee stated plainly during the recorded forum that “a sufficiently malicious customer is, like, basically impossible to prevent at the moment” and that any violations could only be addressed through auditing after the fact or legal action once a contract had been breached. Another participant on the call told the group that CEO Alex Karp “really wants to do this” and that internal efforts to redirect him had been “largely unsuccessful.” The company, they said, was on a “very sharp path of continuing to expand this workflow.”
Thirteen former Palantir employees put their concerns in writing in May 2025, signing a letter condemning the company’s work with the Trump administration. They wrote that when they joined Palantir, they believed in its code of conduct, which stated that its software should protect the vulnerable and ensure responsible development of artificial intelligence. “These principles,” they wrote, “have now been violated.” One former employee captured the shift with precision: “We were supposed to be the ones who were preventing a lot of these abuses. Now we’re not preventing them. We seem to be enabling them.”
The Mirror Palantir Cannot Avoid
External pressure is now matching the internal unrest. The Electronic Frontier Foundation, after sending Palantir a detailed letter asking how its own human rights commitments applied to its ICE contracts, concluded that the company’s responses fell short. EFF argued that a good-faith application of Palantir’s stated principles should lead it to end its work with ICE entirely. Palantir’s response, critics say, relied on polished statements rather than substantive accountability.
The New York City Comptroller’s office, which manages approximately $311 billion in public pension assets, sent a formal letter to Palantir’s board requesting an independent third-party human rights risk assessment of its work with DHS and ICE, citing the stark reversal from the company’s 2020 decision to deliberately avoid contracting with ICE’s enforcement operations over concerns about disproportionate targeting. What had once been a principled line was now a business opportunity.
Palantir’s leadership has shown little sign of changing course. CEO Alex Karp recently told workers the company was “behind the curve internally” when it comes to public popularity, framing employee dissent as a lag in perspective rather than a legitimate alarm. But the workers raising these questions are not behind any curve. They joined a company that promised to stop abuses of power, and they are watching it build the tools that make those abuses faster, wider, and harder to undo. The question they are left with is the one no internal forum can answer: at what point does building the machine make you responsible for what it does?