{
  "record": {
    "id": "0b2b492e-8648-484a-a574-f9f181763199",
    "code": "PAI-0B2B492E",
    "force": "US Immigration and Customs Enforcement",
    "country": "United States",
    "country_code": "us",
    "is_country": true,
    "region": "Nationwide (Enforcement and Removal Operations, Alternatives to Detention programme)",
    "technology": "Hurricane Score (abscondment risk score for Alternatives to Detention)",
    "technology_slug": "individual-risk-scoring",
    "vendor": "Developed in-house by ICE",
    "vendor_slug": "",
    "programme": "",
    "status": "operational",
    "stage": "in-use",
    "evidence": "official",
    "evidence_is_claim": false,
    "since": "February 2019",
    "legal_basis": "",
    "oversight": "Listed in the DHS 2025 AI Use Case Inventory (DHS-2408) as high-impact; the entry records pre-deployment testing, an AI impact assessment, a review by the Chief AI Officer, ongoing monitoring and operator training as in place, and states that an appeal is \"precluded by law\"",
    "notes": "The DHS 2025 AI Use Case Inventory (entry DHS-2408, as consolidated by the Office of Management and Budget) lists \"Hurricane Score\", developed in-house by ICE and operational from 1 February 2019. It is meant to show which noncitizens enrolled in Enforcement and Removal Operations' Alternatives to Detention Intensive Supervision Appearance Program are most likely to abscond. During periodic case reviews an analyst or officer enters information already known about a participant, including case management details and the participant's actions, and the model, described as a quasi-binomial binary classification machine learning model trained on inactive programme case data, returns a score from 1 to 5, higher meaning greater estimated risk. Officers may consider the score with other factors when deciding whether a person's level of case management or technology assignment should change. The entry lists sex and age among the demographic variables the model uses. DHS classes the use as high-impact; it acknowledges that the model can produce false positives that could lead to stricter or more lenient requirements for a person, and says this is mitigated by using the score as one of many factors. The entry records that an appeal is \"precluded by law\" and cites the August 2023 Privacy Impact Assessment for ICE's Alternatives to Detention programme.",
    "source": "Office of Management and Budget, 2025 federal AI use case inventory (DHS-2408)",
    "source_url": "https://github.com/ombegov/2025-Federal-Agency-AI-Use-Case-Inventory",
    "last_checked": "2026-11-06",
    "record_url": "https://www.policeainews.com/tracker/record/0b2b492e-8648-484a-a574-f9f181763199"
  },
  "licence": "https://www.policeainews.com/data#licence"
}