Enhanced Lead Identification and Targeting
Description
The AI is intended to solve the problem of unstructured, hard‑to‑read address information in records like rap sheets and warrants, which makes it difficult and time‑consuming for Enforcement and Removal Operations officers to extract accurate addresses and build usable enforcement leads.
Detailed example
The outputs of Enhanced Leads Identification & Targeting for Enforcement (ELITE) are enriched leads that include AI-extracted addresses. Enforcement and Removal Operations officers review these leads to determine which are actionable and then share actionable leads across offices and areas of responsibility to coordinate enforcement operations.
AI / analytics pattern
Generative AI: AI that generates new or synthetic content (e.g., images, videos, audio, text, code).
Automation level / stage
c) Deployed – The use case is being actively authorized or utilized to support the functions or mission of an agency.
Expected benefit
The integration of AI enhances data extraction capabilities and decreases the time spent on manual data normalization tasks. This provides Enforcement and Removal Operations officers with higher-quality leads and enables them to make better-informed decisions.
Audit / financial statement impact
While ELITE provides actionable data to ERO officers, its outputs are limited to normalized address data and do not serve as a principal basis for decisions or actions with legal, material, binding, or significant effects on individuals. ERO officers review and validate the AI-driven outputs before determining actions, ensuring human oversight and additional verification steps. Furthermore, enforcement decisions are based on the full investigative process, which includes human analysis and validation of the source of AI outputs. As such, the AI system's role is limited to data extraction and normalization, rather than serving as a primary basis for enforcement actions.
Controls / human review
ATO: Yes; PIA: https://www.dhs.gov/sites/default/files/publications/privacy-pia-ice-eid-may2019.pdf
Data needed
The system uses commercially available large language models trained on the public domain data by their providers. The use of LLMs is limited to address extraction from criminal records such as rap sheets and warrants. ICE data was not used during the design, development, or training phases of the AI models. During operation, the AI models interact with ICE production data from multiple sources, including data from ICE’s Enforcement Integrated Database (EID).