DHS · DHS-2388 · record R1887

CBP Translate

Department of Homeland Security · CBP

Find similar use cases DHS profile

Overview

Development stage: DeployedSource: c) Deployed – The use case is being actively authorized or utilized to support the functions or mission of an agency.
High-impact designation: High-impactSource: a) High-impact
Impact justification: Not reported
Start date: 2019-08-07 (day precision)Source: 2019-08-07T00:00:00
Withheld from public reporting: Not reported

Mission

Topic area: Law Enforcement
Operational functions: Translation and transcription Derived by keyword rules; see methodology

Problem the AI is intended to solve

Assist officers and agents with immediate interpretation needs when human translators are not available.

Expected benefits

CBP Translate enhances efficiency by expediting questioning when immediate interpretation is needed. It ensures clear communication, minimizes misunderstandings, and offers immediate accessibility via mobile and web platforms. This improves operational flexibility and creates a smoother experience for travelers.

System outputs

The outputs of CBP Translate include translated text or audio in the form of chat bubbles, which store each interaction. Additionally, CBPOs can capture images of non-travel documents for text translation, but images of actual travel documents are not taken.

Technology

AI classification: Natural language processingSource: Natural Language Processing: AI that processes, interprets, and shares information in human language.
System name(s): CBP Translate
Custom-developed code: Yes
Public source code: Not reported

Sourcing

How it was built: Contract and in-houseSource: c) Developed with both contracting and in-house resources
Vendor (as reported): Aneesh Technologies, 24X7, Ellumen Inc., Deloitte, NiyamIT
Vendors (standardized): Deloitte

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: www.dhs.gov ↗
Authorization to Operate: Yes
Demographic features: Not reported
Training and evaluation data: The models are trained using examples of translated sentences and documents, which are typically collected from the public web. A data miner that focuses more on precision than recall is used, which allows the collection of higher quality training data from the public web.
Federal Data Catalog: Not reported

Governance

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

Pre-deployment testingReported in placeSource: a) Yes

AI impact assessmentReported in progressSource: b) In-progress

Independent reviewReported in progressSource: d) In-progress

Ongoing monitoringReported in progressSource: b) Development of monitoring protocols is in-progress

Operator trainingReported in placeSource: a) Yes, sufficient and periodic training has been established

Fail-safeNot applicableSource: b) Not applicable

Appeal processNot applicableSource: b) Not applicable

User and public consultationReported in placeSource: a) Direct usability testing

Potential impacts and how they were identified

The key risks would be the programs inability to accurately translate what was spoken by both sides of the conversation, leading to significant delays in emergency response situations when trying to leverage traditional phone based translation services in areas with limited cell phone reception. Inaccuracy may also lead to longer processing times at Ports of Entry. These were identified via feedback from the end-users and a common understanding regarding LLM language translation models.