LIGER Generative AI Toolkit

Overview

Development stage

PilotSource: b) Pilot – The use case has been deployed in a limited test or pilot capacity.

High-impact designation

Not high-impactSource: c) Not high-impact

Impact justification

Not reported

Start date

2025-08-26 (day precision)Source: 2025-08-26T00:00:00

Withheld from public reporting

Not reported

Mission

Topic area

Administrative Functions

Operational functions

Summarization, Software development assistance, Translation and transcription, Forecasting and prediction, HR and workforce Derived by keyword rules; see methodology

Problem the AI is intended to solve

LIGER® for FPS will enable FPS users to employ the power of a Large Language Model (LLM) against non-public and sensitive Agency documents to save time and effort conducting time-intensive tasks, such as draft documents to include: Position Descriptions (PD), Statement of Work (SOW) for contracting actions, Professional emails and workforce announcements, Public Affairs stories and releases, and Law Enforcement operations orders; Summarization of large documents; Proofreading and providing feedback or suggestions on written work; conducting Policy analysis to include: Policy comparison and compliance verification, building textual process maps based on policy, and identifying contradictory or outdated policy; Budget forecasting and spend plan analysis; Assisting with code generation, review, and debugging; machine language translation; brainstorming ideas of projects or processes; and information/data retrieval from document libraries. The system will also be used to automate manual processes related to generation of templated documents.

Expected benefits

FPS personnel spend considerable time creating documents from scratch or manually searching large volumes of documents to retrieve information, identify responsibilities, and find discrepancies or outdated passages within policies. Use of LIGER® is expected to substantially reduce time and associated costs in generating new draft documents such as Statements of Work, Law Enforcement operations orders, updated Position Descriptions, and other documents using previous similar examples, reviewing large volumes of information for outdated policy or discrepancies with new DHS policy or Executive Orders, and summarizing large single documents or document collections. LIGER’s ability to securely handle sensitive information and return responses based on custom document collections offers advantages through the ability to securely handle Controlled Unclassified Information (CUI) such as LES and FOUO data which is not suitable for other GenAI applications. Additionally, because LIGER® cites sources, users can rapidly find where specific information from the generated narrative can be found in source documents.

System outputs

LIGER® uses Natural Language Processing (NLP) to return an easily readable text narrative. Like all GenAI, the text response should be verified for accuracy.

Technology

AI classification

Generative AISource: Generative AI: AI that generates new or synthetic content (e.g., images, videos, audio, text, code).

System name(s)

Not reported

Custom-developed code

Yes

Public source code

Not reported

Sourcing

How it was built

Purchased from vendorSource: a) Purchased from a vendor

Vendor (as reported)

LMI Consulting, LLC

Vendors (standardized)

None on the standard list

A vendor is the supplier named by the agency. It does not identify the underlying model or AI technology.

Data and privacy

Involves PII

Yes

Privacy Impact Assessment

Not reported

Authorization to Operate

No

Demographic features

Not reported

Training and evaluation data

LIGER currently uses the ChatGPT-4o large language model provided through OpenAI's service on the DHS Enterprise Cloud - Azure. LIGER itself is not "trained" by the data provided for use with the application and it does not provide user data to LMI or any external entity to fine-tune the components that comprise LIGER. Unlike ChatGPT, LIGER does not have "persistent memory" of inputs provided across different "chat" strings.

Federal Data Catalog

Not reported

Governance

0 of 8 minimum-practice questions answered. A blank answer means the agency reported nothing; it does not mean the practice is absent.

Potential impacts and how they were identified

Not reported