Palantir Has a Human Rights Policy. Its ICE Work Says Otherwise.

Palantir Has a Human Rights Policy. Its ICE Work Says Otherwise. Palantir publishes a human rights policy. They also power the ICE surveillance apparatus. The gap between those two documents is the story. The Human Rights Policy Palantir's Human Rights Policy (updated 2025) commits to: Respecting internationally recognized human rights Avoiding complicity in human rights abuses Conducting human rights due diligence Providing remediation for adverse impacts The policy references the UN Guiding Principles on Business and Human Rights and the International Bill of Human Rights. On paper, it is thorough and commendable. The ICE Work Palantir's systems integrate with: Flock ALPR networks: License plate data flowing from 5,000+ communities ICE deportation databases: Targeting individuals by hundreds of categories including physical characteristics ShadowDragon monitoring: 200+ platforms monitored on behalf of ICE Local police data: Informal sharing agreements with no oversight The ICE database sorts immigration targets by physical characteristics — height, weight, distinguishing features. This is not public safety data management. This is population-scale profiling. The national surveillance network that Palantir helps operate tracks everyone and filters for targets later. The Depth of the Contracts Palantir's entanglement with ICE is not a single contract. It is a decade-long infrastructure buildout. The company's relationship with ICE mirrors its broader government strategy: get the software into one agency, then expand horizontally across the entire federal apparatus. The relationship began in 2011 with a $58,000 server supply deal and escalated rapidly. By 2014, ICE had awarded Palantir the Investigative Case Management (ICM) system — the core software platform that still powers ICE's enforcement operations today. In 2025 alone, ICE paid Palantir $139 million to operate and enhance the ICM system through 2026, then another $30 million for a new platform called ImmigrationOS, followed by a nearly $60 million follow-on contract. The numbers stack up: federal contracts with Palantir have topped $1.9 billion since 2008, with ICE accounting for at least $248 million of that total. In February 2026, DHS awarded Palantir a further $1 billion blanket purchase agreement for AI and data analytics across the department. Palantir's Gotham platform — originally built for counterterrorism analysts — is now the backbone of ICE's domestic enforcement infrastructure. How Flock ALPR Data Feeds ICE Palantir's Gotham platform ingests data from Flock Safety's automated license plate recognition (ALPR) network, which operates across 5,000+ communities nationwide. The architecture is indirect but effective: Flock cameras capture plate data from every vehicle passing through participating neighborhoods. That data is accessible to local law enforcement agencies using Palantir's systems. And because Palantir's ICM and FALCON platforms are the same systems ICE uses, any local police department sharing data on Palantir's infrastructure is effectively feeding ICE's deportation database. The University of Washington Center for Human Rights documented this pipeline in 2025, finding that federal immigration enforcement agencies accessed local ALPR data through "front door, back door, and side door" channels. Audit logs from Virginia jurisdictions showed nearly 3,000 ICE-related searches on Flock data by local agencies. This means communities that installed Flock cameras for neighborhood safety — often sold on promises of reduced auto theft — are unknowingly feeding the deportation machine. Data does not need to flow directly. The system is designed so that local queries and federal queries run on the same graph. Pulling a plate in a sanctuary city on Palantir's platform surfaces connections that ICE can subpoena independently. The data broker economy fills in the rest: Palantir integrates commercially purchased vehicle registration data, utility records, and property ownership databases that have no public safety justification but are invaluable for building deportation target profiles. ShadowDragon and Social Media Surveillance In 2019-2020, ICE awarded Palantir a $49 million contract renewal that integrated ShadowDragon's SocialNet tool into the Gotham workflow. SocialNet monitors over 200 platforms including Twitter, Facebook, Instagram, Telegram, TikTok, Reddit, and niche sites from car forums to dating apps. ICE agents can search social media activity, map personal connections, and build behavioral profiles — all without warrants. A ShadowDragon procurement document obtained by the Brennan Center justified the sole-source contract on grounds that publishing details "would unveil descriptions and techniques used by ICE INTEL that can lead to or tip off person(s) of interest." The tool gives ICE the ability to monitor sentiment analysis, protest organization, and lawful political speech. SocialNet scrapes data from pornographic websites, dating platforms, and tech forums. There is no requirement that targets be suspected of any crime. ShadowDragon's SocialNet and Horizon platforms together query more than 550 data sources through 1,500+ endpoints. ICE is also seeking tools that can "provide monitoring and analysis of behavioral and social media sentiment (i.e. Positive, neutral, and negative)" — meaning it wants to track not just who people are but how they feel about political issues. The capability to monitor protests was confirmed in ShadowDragon's own promotional materials, which cite deployment during political events in Washington, D.C. Legal Challenges The ACLU has documented four Palantir-powered systems driving ICE enforcement: the ELITE targeting application, the ICM database, the ImmigrationOS platform, and the FALCON search tool. In 2026, the ACLU published a comprehensive analysis arguing Palantir's work fails the UN Guiding Principles' complicity standard — the same framework Palantir's own human rights policy claims to follow. The legal challenges are mounting: Warrantless surveillance: ICE agents use Palantir systems to aggregate data from the IRS, Social Security Administration, passport databases, and private data brokers without judicial oversight. A 2026 Colorado federal court ruled ICE had made "unlawful warrantless arrests" and ordered retraining. Family separation infrastructure: Palantir's technology was used in targeted border operations that led directly to family separation and prolonged detention of children in ICE custody, according to a UN human rights filing. Algorithmic bias: The risk-scoring models embedded in Palantir's platforms score deportation targets using criteria that civil rights advocates argue replicate racial and economic bias. The systems tag individuals based on factors like neighborhood, vehicle type, and association networks. A 2017 Freedom of Information Act lawsuit by the Electronic Privacy Information Center (EPIC) forced ICE to disclose documents revealing that the ICM database includes U.S. Citizens — something ICE had not publicly acknowledged. Internal training documents obtained by The Guardian showed ICE agents were instructed on extracting data from phones seized at borders, feeding that data back into Palantir's platforms. The UN Guiding Principles and the Complicity Standard Palantir's human rights policy invokes the UN Guiding Principles on Business and Human Rights (UNGPs). The UNGPs establish that companies must avoid contributing to — or being complicit in — human rights abuses. Legal scholars argue Palantir's ICE work fails this test on multiple grounds. The UNGPs define three forms of complicity: direct (knowingly assisting a violation), beneficial (deriving economic benefit from violations), and silent (failing to speak out when a company has leverage to prevent harm). Palantir's case implicates all three. The company builds the systems ICE uses for warrantless surveillance, knows those systems are central to deportation operations (CEO Alex Karp has publicly defended the work), and profits directly from the infrastructure of enforcement. Amnesty International concluded in a 2020 report that Palantir has a "responsibility to immediately cease their work" with the Trump deportation program — a finding advocates reiterate today. The UNGPs also establish that companies should conduct human rights due diligence that is commensurate with their risk. If a company identifies that its products are being used to facilitate human rights abuses, the responsibility is not merely to document the risk in a policy update — it is to cease the activity or mitigate the harm. Palantir's continued expansion of its ICE contracts after multiple UN filings, ACLU reports, and journalistic exposés suggests the due diligence process is functioning as a compliance checkbox rather than a genuine safeguard. Investor briefings from the Investor Alliance for Human Rights have flagged Palantir's ICE work as a material risk under both human rights frameworks and securities law. The argument: if Palantir's own S-1 filing warned that its AI solutions "are controversial because of their purported or real impact on human rights," and the company continued deepening its ICE relationship, shareholders may have grounds to argue the company failed to disclose known material risks. Palantir's Defense Palantir's standard response is threefold. First, they hold lawful contracts with government agencies and operate within the law. Second, they argue they are not the data controller — they build tools, not the rules for how those tools are used. Third, they point to their human rights policy's provision about conducting due diligence, arguing the policy is live and evolving. These arguments have not held up well under scrutiny. Designing a system that aggregates IRS data, social media activity, and ALPR feeds into a unified deportation targeting interface is a form of policymaking. Deciding which data sources to include, what triggers an alert, and how risk scores are calculated embeds choices that determine who gets detained and who does not. Amnesty International notes that "the architecture of an AI system — how it integrates data, flags individuals, and triggers action — is a form of policymaking." Employee and Shareholder Activism In 2020, activists launched the "Palantir Pledge," calling on tech workers to refuse employment at Palantir until it canceled its ICE contracts. The pledge coincided with protests at Palantir's Palo Alto headquarters involving hundreds of demonstrators from the Coalition to Close the Concentration Camps — an alliance of more than 20 migrant rights organizations. In 2025, thirteen former Palantir employees published an open letter condemning the company's work with the Trump administration. In 2026, former employees further distanced themselves, calling the company's role in mass deportation a violation of the human rights principles Palantir claims to uphold. Investors for human rights groups have filed shareholder proposals demanding transparency about the human rights impacts of Palantir's government contracts. The company's 2020 S-1 filing itself warned investors that "some AI scenarios present ethical issues" and that Palantir may enable "AI solutions that are controversial because of their purported or real impact on human rights." This admission in a securities filing is remarkable: the company told its own investors that its products carry human rights risk — then proceeded to double down on the contracts that create that risk. Protest actions have extended beyond Palantir. Flock Safety has faced city-level pushback, with jurisdictions like Woodburn, Oregon suspending their ALPR programs over ICE concerns, and Culver City, California voting to suspend Flock data sharing with outside agencies. The city-by-city approach may be the most effective lever: if local law enforcement stops feeding data into Palantir's shared graph, the platform loses the breadth that makes it valuable to ICE. The Broader Pattern Palantir is not the only company facing this contradiction. The pattern repeats across the tech industry: Microsoft's HoloLens: A $21.9 billion contract with the U.S. Army for augmented reality combat headsets, while Microsoft publishes its own responsible AI principles. Amazon Rekognition: Before a 2020 moratorium, Amazon sold facial recognition to law enforcement, and the ACLU demonstrated it falsely matched 28 members of Congress with mugshots. Google's Project Maven: Google cancelled the drone imagery AI contract after employee protests, but Palantir stepped in and now runs the program. Clearview AI: Scraped billions of facial images from social media without consent, sold the database to ICE, and inked a $9.2 million contract with the agency in 2025. The common thread is that each company publishes values on one page while signing contracts on another. The revolving door between government and corporations ensures that regulation of these contracts is designed by the people who benefit from them. And Anthropic's partnership with Palantir proved that even companies founded on AI safety missions will make deals with surveillance infrastructure once the valuation gets high enough. These are not isolated ethical lapses. They are the normal operation of an industry whose incentive structure rewards government revenue over human rights commitments. The companies that say no — Google abandoning Project Maven, Amazon pausing Rekognition sales — are notable precisely because they are exceptions to the rule. The Gap Palantir's policy says they avoid complicity in human rights abuses. Their ICE work enables warrantless tracking, family separation, and deportation based on algorithmic risk scores trained on biased data. The documents are irreconcilable. One of them is marketing. Every tool you use — every search, every data point, every logged plate — feeds systems that can be turned against you. The only protection is not to need protection. Use Signal. Use Tor. And treat any device that crosses a border as compromised the moment it leaves your hands.