CBP_AIIndustryDay_Slides_5.29.24.pptx

May 29, 2024

AI Industry Day Agenda Opening Remarks AI at CBP Mission Panel Discussion AI and Procurement at CBP Closing Remarks

May 29, 2024

Opening Remarks:

James McCament, Acting Chief Operating Officer

Artificial Intelligence at CBP

AI at CBP Presented By Joshua Powell, Acting Director of AI Implementation, Office of the Commissioner

Trang Tran, Deputy Chief Data Officer, Director of Artificial Intelligence Center of Innovation (COI) Office of Information & Technology (OIT)

Sunil Madhugiri, Chief Technology Officer, Office of Information and Technology (OIT)

AI Overview

What is Artificial Intelligence according to CBP?

Any artificial system that performs tasks under varying and unpredictable circumstances without significant human oversight, or that can learn from experience and improve performance when exposed to data sets.

Machine learning Enables computers to learn from examples and experiences and make decisions on its own, based on given data.

Optical character recognition (OCR) Teaches computers to read and understand text from images or scanned documents and automatically convert into digital formats.

Natural language processing Helps computers understand, interpret, and generate human language to enable technology like virtual assistants and text analysis.

Computer Vision Allows computers to recognize objects, people, text, images, and video to perform tasks such as object detection and facial recognition.

What are the primary subsets of AI?

The History of AI at CBP

CBP begins adoption of AI through use of machine learning on Targeting Models

Modeling and Simulation tools are developed and tested for USBP and OFO 2014-2016 AI Facial Recognition is piloted through Traveler Verification Service AI Computer vision technology is piloted via Autonomous Surveillance Towers Small Unmanned Aircraft Systems (sUAS) and Tethered-sUAS leverage computer vision for detection and collision avoidance

Natural Language Processing is used to validate identities within the CBP One App Computer vision capabilities automatically detect items of interest in video 2021-2022 Facial biometrics and color illumination are leveraged to determine traveler status 2021-2022 Autonomous navigation and maritime detection systems are piloted Three-dimensional images of occluded objects are produced and autonomously identified

CBP has used AI to support the Mission for over a decade:

In 2024 and beyond, CBP will prepare the Mission for Generative AI

May 29, 2024 AI Policy and Guidance Implementation at CBP In support of CBP Priority 9.02 and the OMB memorandum (memo) establishing requirements for AI governance, INVNT, Office of Information Technology (OIT), Chief Data Officer (CDO), and Chief Technology Officer (CTO) initiated an inventory of AI related use cases at CBP ahead of Department of Homeland Security (DHS) schedule.

Executive Order 13960 (December 2020): Promoting the Use of Trustworthy AI in the Federal Government

Executive Order 14110 (October 2023): Safe, Secure, and Trustworthy Development and Use of AI Policy Statement 139-06 (August 2023): Acquisition and Use of Artificial Intelligence and Machine Learning by DHS Components

Policy Statement 139-07 (Oct 2023): Use of Commercial Generative Artificial Intelligence Tools OMB Memorandum (March 2024): Advancing Governance, Innovation, and Risk Management for Agency Use of Artificial Intelligence

Executive Orders (EO) DHS Policy OMB Guidance

As directed by the authoritative documents outlined below, it is critical for CBP to capture an accurate AI Use Case Inventory to bring transparency and understanding around each CBP AI effort.

AUTHORITATIVE DOCUMENTATION

May 29, 2024 Air and Marine Operations What are the top three challenges facing your component that might be addressed by AI?

The operation of unmanned aircraft systems within airport traffic areas, at national security special events, and for nefarious purposes (collection, observation, trafficking) at the border and in proximity to security operations is a national security threat. AI may be useful in the timely detection and classification (friend/foe) of UAS at/near sensitive locations and in the determination of effective countermeasure options for threat mitigation.

Thousands of aviation tracks over and near the United States are collected by sensors and analyzed by human controllers who attempt to use years of experience to identify illicit activity including possible terrorist threats and smuggling. With the expected increase in unmanned aviation, the numbers of targets to track are expected to increase dramatically. Additionally, unmanned activity will not be limited to a limited number of airfields. AI may be useful in the identification of behavior consistent with illicit activity allowing human observers to focus on AI-cued aircraft tracks.

Commercial satellite imagery has become more available and often collects on the maritime environment. AI may be useful in the use of commercial satellite imagery to track maritime activity and identify behavior consistent with illicit activity allowing human observers to focus on AI-cued maritime tracks.

How can industry help your organization address these mission challenges?

AMO depends on industry and government R&D to provide the tools necessary to assist in efficient and effective safeguarding of our nation through anticipation and confrontation of air and maritime security threats.

May 29, 2024 Office of Field Operations What are the top three challenges facing your component that might be addressed by AI?

Use Case 1: Technology that conducts multimodal targeting and analysis of de minimis shipments at the tactical level.

Use Case 2: Technology that takes metadata from derogatory activities and searches for similar activities across the entire dataset.

Use Case 3: Technology designed for the administrative optimization of business processes.

The advancements in machine learning and artificial intelligence have enabled the achievement of more sophisticated data analytics objectives, deeper targeting strategies, and the streamlining of administrative functions.

May 29, 2024 Office of Trade What are the top three challenges facing your component that might be addressed by AI?

Identify Intellectual Property Rights (IPR) violations by providing users with current counterfeit methodologies and/or automatically comparing a product to an image and alerting users to discrepancies.

Identify forced labor violations and illegitimate trade in the supply chain.

Identify money owed to the government by aiding research and analysis, continuous monitoring to track compliance, improving data collection and organization of both structured and unstructured data, and enhancing fraud detection risk assessments.

How can industry help your organization address these mission challenges?

Submit AI solutions to OT through Business Connections CBP Business Connection Expanding Its Reach - All U.S. Companies Are Now Eligible to Register | U.S. Customs and Border Protection Join OT working group(s) (e.g., COAC) Commercial Customs Operations Advisory Committee (COAC) | U.S. Customs and Border Protection (cbp.gov) Industry could create low level AI programs (limited memory Ai, Generative Ai, and Natural Language Processing Ai) that require smaller footprints, less specialized knowledge and can operate across enterprise libraries simultaneously for discrete CBP workplace processes aligning to S1’s goals to build an Ai-enabled workforce, cohesive partnerships, and shared AI Infrastructure.

May 29, 2024 U.S. Border Patrol What are the top three challenges facing your component that might be addressed by AI?

U.S. Border Patrol agents often face a multitude of challenges, but three of the most common ones that might be addressed by AI include:

Use Case 1: AI for enhanced domain awareness, analysis, and tracking.

Leveraging AI can improve capabilities to detect, identify, classify, and track multiple items of interest across different domains Use Case 2: AI for operational efficiency and resource allocation U.S. Border Patrol can leverage AI to automate routine administrative tasks Use Case 3: AI for predictive analytics and proactive management How can industry help your organization address these mission challenges?

Help the USBP solve challenges with automated data sharing challenges as well as computer vision solutions that would reduce the need for additional agents to monitor domain awareness applications.

Help the USBP find novel ways to rapidly conduct a search of data sources, and the ability to scrutinize, distill, label, and preserve data submitted by staff utilizing a variety of resources.

Supporting the Check Point of the future with artificial intelligence and additional capabilities to create efficiencies in USBP Non-Intrusive Inspection.

May 29, 2024 Operations Support What are the top three challenges facing your component that might be addressed by AI?

CBP Planning, Analysis, and Requirements Evaluation Directorate (PARE) lacks a machine interpretable layer for the human understandable aspect of their data. Knowledge graphs enable this interface as part of data-centric architectures. PARE needs automated, application-agnostic, and scalable solutions that use existing semantic standards to produce AI-ready data at mission speed.

CBP Laboratory and Science Services (LSS) is seeking to use AI technology to enhance their ability to determine the source of origin (geographic or cartel-specific), location, and method of synthesis for fentanyl and other synthetic narcotics.

CBP International Affairs (INA) is seeking AI-powered translation tools to help bridge language barriers by quickly translating documents, remarks, and meeting notes in real-time to facilitate international communication and negotiations.

PARE: Recognize the requirements for non-proprietary tool-agnostic standards for semantic technology, applied ontology, and knowledge graphs. (DoD/IC/ISO Standards)

May 29th, 2024 Enterprise Services

What are the challenges facing your component that might be addressed by AI ?

Human Resource Management (HRM): Significant workload to retrieve and interpret policy, employee data, and case histories to answer personnel inquiries (e.g. retirement question, position descriptions). Additionally, CBP sees a large amount of taskings from the hill, industry, and the public.

(HRM): Significant workload to provide personalized customer service to CBP recruits and applicants. This work includes resolving common queries and issues to keep customers informed and streamline the workflow.

Office of Acquistion (OA): Complex analytics to predict acquisition program and portfolio outcomes based on program data (e.g. cost, schedule, funding, technical performance, risks, politics, etc.)

(OA): AI-enabled system to search a database of historical and open sources that meet a set of user specified requirements, matrix results (linked to the source documents) and identify additional criteria that discriminate between products, reducing time/burden.

Office of Training and Development (OTD): Creating educational content from review of unfamiliar complex documents (e.g. executive orders, curricula, classroom presentations) Digestion is labor-intensive and time-consuming.

May 29th, 2024 Enterprise Services - OIT What are the top three challenges facing your component that might be addressed by AI ?

Cybersecurity Directorate (CSD) Use Cases: Enabling CBP users to leverage AI while also maintaining and complying with security, privacy, legal, and other requirements. Ensuring AI leveraged is free from bias, provides accurate and legally defensible responses/information, and can be protected from vulnerabilities and adversarial attack/exploitation Passenger Systems Program Directorate (PSPD) – Use AI to assist in automating initial categorization and responses to incoming feedback/emails from mobile users.

PSPD - Automation of Currency and Monetary Instrument Report (CMIR) forms – Use AI to extract handwriting from scanned CMIR forms and automatically populate necessary systems based on quality / confidence thresholds.

PSPD - Automation Development – Use AI to spawn a website/application and find all permutations and combinations for data related fields and then use that data for testing edge cases, negative scenarios.

Targeting Analysis Systems Program Directorate (TASPD) Use Case : Leveraging AI to reduce false positives in our targeting systems (referrals with no action required) and improve true positives (referrals with action required, whether seizure or denial of entry). AI presents an opportunity to improve effectiveness (true positives) and efficiency (false positive suppression).

How can industry help your organization address these mission challenges?

Aid the government in establishing standards, guidance, policies, and procedures for safe and secure use of AI Don’t just offer products/companies as an “AI” vendor. Provide details on what specifically the offering does with AI and why the product/service is better than their product that does not or did not use AI

Industry Partnership

May 29, 2024 AI Center of Innovation (AI COI) Data Ecosystem AI Data Library: repository of annotated CBP data w/ functional data pipeline

AI Data Interface: easy to use search interface Data Annotation Data Enrichment

AI Community Education: empower CBP community with skills, resources, and knowledge to leverage AI ethically and effectively

Engagement & Governance:

Foster a community of learning and excitement around AI and enterprise understanding of the portfolio

Industry Partnership: Enable our industry partner community with sanitized training data (public release approval) so they can bring advanced AI solutions to CBP

CBP AI CENTER OF INNOVATION (COI) Vision Rapid AI Solutions Use Cases: creation of AI solutions for the toughest mission problems Tiered AI Solutions Service Model: providing AI data, services and tools 3rd party annotation services: prepare data for AI Enrichment services: prepping data (including legacy) for annotation

Data Pipeline How we move data into and out of the platform Model Pipeline How do we manage the model lifecycle from development through deployment

Data Platform What data management mechanisms are needed to store and share data The AI Center of Innovation and Industry A foundational ecosystem that supports partnership between industry and CBP to provide access to technologies, processes, and tools that support data enrichment, annotation, and AI solutions that can be scaled for the CBP enterprise.

May 29, 2024 AI Ops: Operationalizing AI Solutions What is “AI Ops” Managed technology environment with simplified workflows to build and test AI models, provide security reviews, and deploy models into operation.

Like our software development approach, OIT leverages streamlined processes and proven methods for the development of the AI Ops Environment.

Streamlines the end-to-end AI operations life cycle from data collection and processing, to training and tuning, and through deployment and monitoring.

BENEFITS

METHOD

Increased Responsiveness Improved Flexibility Faster Deployments

AGILE DEVELOPMENT

Improved Data Quality Faster Integration Data Standardization

DATA PIPELINES

Minimized Risks Enhanced Reliability Improved Decision Making

GOVERNANCE

Lower Costs Increased Scalability Amplified Security

COMMON SERVICES

Generative AI at CBP

May 29, 2024 Generative AI at CBP

ARTIFICIAL INTELLIGENCE (AI)

Machine Learning (ML) Deep Learning (DL) Generative (Gen AI) Natural Language Processing (NLP) Computer Vision

(CV)

Imagery / Video Multimodal Text / Audio

TYPES

DOMAINS

MODALITIES

Natural Language Processing (NLP) is a subfield of artificial intelligence focused on enabling computers to understand, interpret and generate human language.

Information Retrieval Language Comprehension Emergent Capabilities

Gen AI Demystified Large Language Models (LLMs) are deep learning algorithms notable for general-purpose language understanding and generation.

is a specialized branch of artificial intelligence comprised of deep learning models that empower AI systems with creative content generation, complex data understanding, and probabilistic decision-making.

Generative AI: Overview

A Gen AI Approach:

Cloud Services and Open-Source AI By combining Open-Source AI models with cloud services, CBP can take mission-critical AI projects beyond the proof-of-concept phase with scalability, security, flexibility, and cost-effectiveness in mind.

Cloud Services and Open AI Models

Open-Source AI Model Examples

Cloud A Cloud Service AI Models

LLM

Cloud B

Cloud C

Retrieval Augmented Generation (RAG) to access and search a defined corpus of internal data Optimized responses with retrieval citations Secure connection to internal data sources Approach Outcomes for CBP End User

A CBP Gen AI Use Case: LLM Policy Bot

An LLM Use Case A securely-deployed LLM-enabled tech stack intended to provide an NLP search system to retrieve relevant CBP information.

Benefits to CBP

Panel Discussion:

AI in the Mission at CBP

Panel Discussion:

AI In the Mission at CBP Moderated by:

Ntina Cooper, Deputy Executive Assistant Commissioner, Enterprise Services Panelists:

Sanjeev “Sonny” Bhagowalia, Assistant Commissioner, Office of Information & Technology (OIT) and CBP Chief Information Officer (CIO) Jody Hardin, Executive Director, Innovation and Strategy Directorate, Office of Field Operations (OFO) Henry Laxdal, Deputy Executive Director Program Management Director, U.S. Border Patrol (USBP) James Byram, Executive Director, Trade Transformation Officer, Office of Trade (OT) James Thom, Acting Executive Director, Domain Awareness Security Operations, Air and Marine Operations (AMO)

May 29, 2024 AI In the Mission Panelists Photo Here Sonny Bhagowalia Assistant Commissioner, Chief Information Officer, Office of Information & Technology (OIT) Photo Here Jody Hardin Executive Director, Innovation and Strategy Directorate, Office of Field Operations (OFO) Photo Here Henry Laxdal Deputy Executive Director Program Management Director, U.S. Border Patrol (USBP) Photo Here James Byram Executive Director, Trade Transformation Officer, Office of Trade (OT)

Photo Here James Thom Acting Executive Director, Domain Awareness Security Operations, Air and Marine Operations (AMO) Photo Here Ntina Cooper Deputy Executive Assistant Commissioner, Enterprise Services

AI and Procurement at CBP

AI and Procurement at CBP Presenters Joshua Bedregal, Branch Chief, Mission Support Contracting Division, Office of Acquisition Nick Martinelli, Director, Information Technology Contracting Division, Office of Acquisition (OA) Dyad Searcy, Procurement Analyst, Office of Acquisition (OA)

Example Procurement at CBP

NII ADA Procurement

PURPOSE

CSOP

Utilized Commercial Solutions Opening Pilot Program (CSOP) for the acquisition process of procuring Artificial intelligence for Anomaly Detection Algorithms (ADA)

ADA PROCUREMENT PHASES

Integrate Artificial Intelligence/algorithms into CBP’s Non-intrusive Inspection (NII) systems for contraband detection and anomaly screening.

Leverage industry produced solutions to increase effectiveness and efficiency of NII operations.

Each contract has a maximum ceiling of $25M.

Provides procedural discretion and flexibility for efficient acquisitions.

Merit-based solution selection strategy derived from NDAA for FY 2017, (amended in Dec. 2023) allowing streamlined and less restrictive acquisitions of innovative commercial items.

CSOP BENEFITS FOR NII ADA:

Collaborative terms negotiation with industry (e.g., data rights terms).

Flexibility to onboard additional vendors.

Flexible for traditional and nontraditional vendors

Phase 1: Short White paper submission and Optional technical demonstration.

Phase 2: Request for full proposal with price and terms negotiation.

Acquisition Innovation Advocate’s AI Procurement Options For the Future CBP will ensure to leverage the appropriate acquisition strategy that best meets the Government’s needs.

Future AI Procurement Strategies to Consider:

Traditional FAR acquisitions

Small Business Innovation Research (SBIR)

Commercial Solution Opening Pilot Program (CSOP)

New Solutions Marketplace, as of 2022:

Tradewinds:

Digital repository of readily awardable AI capability videos.

Addresses Government’s challenges in AI/ML, digital, and data analytics.

CBP In-House Procurements

May 29, 2024 CBP In-House Procurements CBP purchases AI through the Office of Information and Technology Software Application and Services software development contracts for processing and enforcement.

Targeting and Analysis Systems Program Directorate (TASPD) Passengers Systems Program Directorate (PSPD) Cargo Systems Program Directorate (CSPD) Border Enforcement and Management Systems Program Directorate (BEMSD)

CBP can utilize internal agency vehicles to purchase AI.

Enterprise Small Business Enterprise Business Management and Support Services

AI Procurement Policy

May 29, 2024 AI Procurement Policy March 1, 2024, DHS Acquisition Alert 2024-06: Information and Resources related to the Procurement of Artificial Intelligence Technologies at DHS and Evolving Regulatory Landscape

April 4, 2024, CBP Procurement Directorate established an “AI Policy Team”

May 21, 2024, CPO Memo: Emerging Technologies (ET) Deployment in Contracting Activities

Closing Remarks

Closing Remarks Provided by Julie Koo, Director of Industry Partnership & Outreach Program, Office of Acquisition (OA)

Please submit questions through Business Connection:

https://cbpbusinessconnection.cbp.dhs.gov/vmp?id=vmp_vendor_signup image1.jpeg image2.jpeg image3.jpeg image5.png image4.jpeg image6.png image7.svg

.MsftOfcThm_Accent1_Fill_v2 { fill:#156082;

image16.png image17.png image8.jpeg image9.jpeg image10.png image11.jpeg image12.jpeg image13.png image14.png image15.jpeg image26.svg

.MsftOfcThm_Background1_Fill_v2 { fill:#FFFFFF;

image18.png image19.png image20.png image21.png image22.svg

.MsftOfcThm_Background1_Fill_v2 { image23.png image24.svg

.MsftOfcThm_Background1_Fill_v2 { image25.png image27.png image28.png image29.png image30.png image31.png image32.png image33.png image34.png image35.png image36.png image37.png image38.png image39.png image40.png image48.svg

.MsftOfcThm_Background1_Fill_v2 { image41.png image42.svg

.MsftOfcThm_Background1_Fill_v2 { image43.png image44.svg

.MsftOfcThm_Background1_Fill_v2 { image45.png image46.svg

.MsftOfcThm_Background1_Fill_v2 { image47.png image56.jpeg image57.jpeg image49.png image50.svg

.MsftOfcThm_Accent1_Fill_v2 { fill:#156082;

.MsftOfcThm_Accent1_Stroke_v2 { stroke:#156082;

image51.png image52.svg image53.png image54.jpeg image55.png image65.png image66.svg image67.png image68.svg image58.png image59.svg image60.png image61.png image62.svg image63.png image64.svg image69.jpeg image70.jpeg image71.png image72.jpeg image73.jpeg image74.png image75.png