Emergency Hotline: Call 1-844-363-1423 (United We Dream Hotline)
ICE Encounter
Emergency

Data Analytics & Algorithmic Enforcement

Documentation of AI and data analytics in immigration enforcement including Palantir ICM, ImmigrationOS, Risk Classification Assessment algorithm, and social media monitoring programs.

Overview

The volume of data collected by DHS far exceeds human analytical capacity. ICE has contracted with defense technology firms to build algorithmic platforms that automate targeting, detention decisions, and deportation logistics.

Palantir ICM

Contract Overview

Since 2014, Palantir Technologies operates the Investigative Case Management (ICM) system as ICE's official system of record.

Contract Element Value
Base ICM contract $115M+
ImmigrationOS expansion (2025) $30M
Total $145M+

System Capabilities

ICM is not a passive database—it's an active predictive intelligence tool.

Data Fusion

ICM cross-references data from:

  • FBI
  • DEA
  • ATF
  • Student and Exchange Visitor Information System (SEVIS)
  • Commercial records
  • License plate readers
  • Mobile devices seized at border

Dossier Building

Agents use ICM to construct 360-degree profiles linking:

  • Biometric data
  • Biographic information
  • Travel history
  • Financial records
  • Social associations
  • Location history

ImmigrationOS

Contract Award

In April 2025, ICE awarded Palantir a $30 million sole-source contract for ImmigrationOS.

Justification

The sole-source justification cited Palantir's "deep institutional knowledge of ICE operations."

AI Capabilities

ImmigrationOS leverages artificial intelligence across three functions:

Function Description
Targeting & Prioritization Algorithmically identify and prioritize apprehension targets
Real-Time Monitoring Track visa overstays and "self-deport" subjects
Logistical Optimization Streamline deportation from identification through removal

Target Categories

System focuses on:

  • "Violent criminals"
  • Suspected gang members
  • Visa overstays
  • Self-deportation non-compliance

Delivery Timeline

Prototype scheduled for September 2025.

Risk Classification Assessment (RCA)

Purpose

The RCA algorithm (introduced 2012) determines whether apprehended migrants should be:

  • Detained
  • Released on bond
  • Released on own recognizance

Original Intent

Promoted as objective reform to:

  • Rationalize detention procedures
  • Reduce unnecessary incarceration
  • Base decisions on calculated risk

FOIA Revelations

Extensive disclosures reveal the RCA was subverted to maximize detention.

Manipulation Tactics

Tactic Effect
Subjective input manipulation Officers gamed inputs to ensure detention
Business rule changes (2015) Removed ability to recommend bond
Business rule changes (2017) Virtually eliminated "release" output

Release Rate Collapse

Year Algorithmic Release Recommendation
Pre-2017 ~10%
Post-2017 <5%

Academic Assessment

Scholars characterize the RCA as providing a "veneer of risk" and procedural legitimacy to what functions as automated mass incarceration.

Social Media Monitoring

Scope

DHS extensively monitors social media to:

  • Gather intelligence
  • Vet immigration benefit applicants
  • Generate investigative leads

Contractors

Contractor Capability
Giant Oak Social media analytics
Shadowdragon Network mapping and scraping

Programs

Manual Review

CBP and ICE conduct manual reviews of publicly available profiles.

Automated Scraping

Contractor software:

  • Scrapes public social media content
  • Maps social network associations
  • Flags content for review

Effectiveness Questions

Brennan Center Review

Comprehensive analysis found:

Out of thousands of immigration benefit and refugee cases vetted through DHS pilot programs, social media screening failed to yield "clear, articulable links to national security concerns."

Conclusion

The practice demonstrates:

  • Lack of data integrity
  • Massive expansion of digital footprint monitoring
  • Negligible security benefits

Algorithmic Bias Concerns

Training Data Issues

Algorithms reflect biases in their training data:

  • Historical enforcement patterns
  • Existing demographic disparities
  • Geographic targeting decisions

Lack of Transparency

  • Proprietary algorithms are not publicly disclosed
  • No independent audit of decision criteria
  • Limited ability to challenge algorithmic determinations

Human Accountability

Algorithmic systems:

  • Strip human accountability from enforcement decisions
  • Hide punitive policies behind "objectivity"
  • Create difficulty in legal challenges

Documentation Sources

FOIA Releases

Significant ICM documentation obtained through:

  • ACLU litigation
  • Electronic Frontier Foundation requests
  • Academic researchers

Contract Records

Available through:

  • USAspending.gov
  • Federal procurement databases
  • Congressional oversight documents

Implications

For Individuals

Understanding algorithmic enforcement helps:

  1. Recognize that detention decisions may be algorithmic
  2. Understand data sources feeding targeting systems
  3. Assess exposure from various data footprints

For Advocates

Documentation supports:

  1. Challenging algorithmic detention decisions
  2. FOIA requests for specific system documentation
  3. Policy advocacy for algorithmic transparency
  4. Litigation on due process grounds

For Attorneys

In detention cases:

  1. Request documentation of RCA score and inputs
  2. Challenge algorithmic determinations
  3. Document input manipulation patterns

Related Resources