Innovation in a Dynamic Trade Environment: CBP’s Trade Enforcement Mission Part 2

Welcome to “HSDF THE PODCAST,” a collection of policy discussions on government technology and homeland security brought to you by the Homeland Security and Defense Forum

 Goods don’t talk, but the data around them does—and that’s where the real border work begins. We take you inside how Customs and Border Protection secures modern commerce by pairing AI with seasoned human judgment to spot risky shipments, speed trusted trade, and protect consumers and industries from harm.

 Featuring:

• Brandon Lord, Executive Director, Office of Trade Policy & Programs, CBP
• Eric Choy, Executive Director, Trade Remedy Law Enforcement, Office of Trade, CBP 
• C. Shane Campbell, Executive Director (Acting), Cargo and Conveyance Security, Office of Field Operations, CBP
• Brenda Smith, Former EAC for Trade, CBP (moderator)

 This discussion took place December 12, 2025 at 8th Annual Homeland Security & Defense Forum Border Security Symposium

Follow  HSDF THE PODCAST and never miss latest insider talk on government technology, innovation, and security.  Visit the HSDF YouTube channel to view hours of insightful policy discussion.  For more information about the Homeland Security & Defense Forum (HSDF), visit hsdf.org.

Transcript

Framing The Data Landscape

SPEAKER_03 0:00

To um give another facet of um the environment around the data. So Eric talked about all of the data that is looked at. But in your world, um, and I know the the um when uh deputy commissioner talked about the National Targeting Center, um there has been great synergy between the anti-terrorism mission and the trade mission. The connection is the data. Can you talk a little bit about the data that CBP collects related to shipments and how it's used by CBP and how that impacts kind of trade security or economic security? And and maybe, just to throw in, there there was a he addressed the de minimis concept. So maybe you could talk a little bit about the specific challenges there.

AI’s Limits And Human Judgment

CTPAT And Accelerating Reviews With AI

Scaling Anomaly Detection At The NTC

SPEAKER_01 1:04

Oh wow. Oh, I see Brandon's giving me the look of this because he knows how much I love talking about this topic. So I'm gonna take the data, I'm gonna build off of Eric's thing, and I may or may not answer your question directly and hold me to that if I don't, right? Because I I do want to go back and talk about the use of AI and hit on something that Eric um touched on briefly, right? Which is, you know, time and time again, um, and I'm gonna caveat this to where I'm a fan of AI, but I'm a fan of AI up to a point, right? Because we we have a notion or when we talk about AI, it makes people nervous because it's, you know, it's gonna replace humans, it's gonna replace workers, et cetera, et cetera. And I can tell you in my line of work, no. Because test after test after test has shown that when you take the human capital part out of the AI and test us as people, we're much better and we perform much better, right? Now, is there a chance that 20 years down the road that AI can get better? Okay, maybe, but AI just doesn't have the same synapses and stuff that we have in our brains and therefore can't do the reasoning, et cetera, et cetera, that we get. Correct. So, well, I mean, this goes back to complex, right? Which was in its way a light form of AI of just doing random examples, right? And it's still an algorithm, although a very simplistic algorithm. Um and time after time, over, I mean, we've been doing complex exams, I've been in 30 years, and it's way longer than 30 years that CBP officers are much better than the technology. So I'll just hit that one right there. Um, when it comes to data, I'm gonna turn your example to something that may be um, at least in my head, easier to explain. We have a CTPAT program, Customs Trade Partnership Against Terrorism, with 11,000 members. Uh, and those members um kind of guarantee that they are protecting their supply chain. It's a very rigorous application process. In fact, the application for us to do one new CTPAT partner applying for the program, it's about a four to five month process for us to go through a validation process. And the biggest time constraint on that is going through the hundreds, if not thousands, of pages of documents that are presented to our supply chain security specialists who are reviewing those companies, going through their books, and really getting into the weeds of the business operations of that company. So when you have an individual who's got to go through thousands of pages of documents looking to make sure that every question was answered, that commas aren't out of place, that you know the appropriate information matches the question, or that the question has actually been actually answered and not just fluff trying to fill in words to make it look like questions were answered, right? We've done testing with AI showing that AI can help us do that clerical review process at a rate of 60% faster than what our supply chain security specialist can do on its own. So it really can go through and just flag those particular areas that we really need to look at and dig deeper into because questions weren't answered or just omitted on its face, right? That saves tens, twenties, hundreds of hours of time that our supply chain security specialists don't get stuck reviewing paperwork, but they can actually go out and do physical validations of the companies and their supply chains that better protect the supply chain of the United States, right? So if you if you take that and link that in with the I can't like billions of bytes of data that the NTC is dealing with on a day-to-day basis, you can bring that concept up to that level too, where AI is helping with our algorithms to look for those anomalies and targeting that to AI looks weird, but then the we officers, other specialists step in to really look at that and to dig a little deeper to make sure is it really weird or is it something we truly have to be worried about? Does that answer your question?

Pre‑Arrival Data And Risk Targeting

SPEAKER_03 4:59

Um it gets there, it gets there. So um you collect, we, sorry, CBP collects um data in advance of a shipment actually arriving, and you're doing reviews at that point. Um goods actually arrive, and you're looking again to make sure there's no bomb in that box or that there are no counterfeit tennis shoes in that shipping container, and then you do reviews after. My sense is that the tools that CBP is using to check the algorithms, to flag the risky shipments, so we know which uh shipping containers or small packages really need to be opened up, um, or perhaps X-rayed or run through a radiation portal monitor is really important. But what I'm seeing over time as CBP develops new tools is the expectations for the private sector on applying those tools. You have anything you want to add? And Brandon, I'm gonna give you a chance maybe to build on what your colleagues have said and then yeah, sure.

Industry Expectations And Shared Tools

Open Source Data And Tracing People

Shell Companies And Evasion Tactics

SPEAKER_02 6:17

So, you know, Shane's absolutely right about you know the use of AI and our CTPAT program. You know, the unique challenge in in the customs space and the trade space is you know, we're we're trying to do two things. We're trying to protect our national economic security from imports that unfairly undercut or harm our domestic industries, and we're trying to keep fake counterfeit or unsafe products out of the out of the out of the country from entering the country. The problem is, the the core challenge is all of that harm, all those, all that activity is is frankly finalized overseas and then sent here. And then when the harm gets here, it's a thing that doesn't talk to you. It's a it's a it's a product. And so it doesn't look nervous when you interview it. It can't tell you where it traveled in the last six months, right? I'm not an immigration guy, right? But like we've all sort of processed through an airport, right? We can't do that, right? So being able to take the data that we get pre-arrival or the data that we get through open source, and being able to link it in new ways to the people behind these activities further helps our mission. And so what we're saying to industry now is hey, there is a wealth of open source data that has never existed uh at this scale now, right? To say that you no longer know is increasingly not an acceptable answer. Uh at the same time, though, we're also asking industry, what sort of industry standards or tools that you are using to answer these questions that we can potentially leverage? We very much don't want to get into a scenario where, as a customs, as a customs authority, we're imposing some sort of like a traceability, you must do it exactly this way, standard, or some other sort of like data format or tool that you must adopt and take on costs. What we're asking broadly is what tools are out there for supply chain tracing? What tools or standards are you using to better understand and see into that supply chain to better understand origin? And oh, by the way, we're probably gonna want to increasingly know a little bit more about who the people are behind these companies. Because one thing we haven't talked about is shell companies. And that's that's a whole nother challenge. And the shell company problem is super simple. Very easy to establish companies in the United States with limited information. Those companies are set up for one purpose to evade US trade laws. Okay? And people think, oh, this should be easy. It's not easy because they pay the money that we that they're owed, that they owe us. Right? It would be easy if they just skipped town and didn't pay. They don't skip town until we find them, right? And then when we find them and realize they owe us money, now they're gone. But the truth is they've probably set up other shell companies and they're just waiting to show up as a brand new company. Right? So data and you know, verifiable credentials, you'll hear some people talk about that, but really like the data, knowing the people and the and the entities that are behind these companies, what they're really doing overseas, what processes they're leveraging overseas, all that data combined with open source data is going to increasingly help us. Because again, you know, imagine you open up your, you know, your box shipped to your door of whatever you or your spouse or your kids ordered, and you had to find out from that box, like, where exactly were you manufactured? Right? The box won't tell you anything, right? And so our officers and our import specialists, like that's what they have to work with, plus whatever data we can get them, right? And that that's like that's the core challenge of the time, right? That is that is what we're faced with today.

De Minimis Changes And Mail Data

Magic Wand: Traceability And AI Fusion

SPEAKER_01 10:05

Brenda, I'm gonna add in because you mentioned Deminimus in your last question, and I tried to do the headquarters thing of going around it, right? But it it plays into what Brandon's saying, and de minimis is a perfect example of that, right? Because this was an environment for those of you not familiar with DeMimis that allowed uh the passing of anything under$800 coming into the United States to declare duty-free, essentially, right? And that changed just a few months ago. And when the announcements came out, you would have thought that the world was gonna end, right? And and that was the consternation that we were getting from trade industry, from carriers. Um, name your field across the cargo and trade environment. And and and there was a lot of consternation about the end of this. The world didn't stop spinning. Because as our push to get more data um from our trade partners, you know, they they did it. They're they're coming back, right? Mail is a perfect example of that. Where um, and I'm gonna say this too, like in the de minimis environment, that was like the bulk of our seizures. It was something like 90 plus percent of our seizures by counts happened in the de minimis environment. There were some true threats there because we weren't getting the information and data that we could use to pre-screen and to pre-clear some of this stuff that was coming through. And when you're talking about over 4 million shipments a day coming into the United States, that's a lot of stuff to look at. That's a lot, right? Um, and usually when we talk about de minimis, a lot of folks focus on mail, right? Mail was only 5% of the de minimis volume. That and that was shocking to some folks, you know, who learned that mail wasn't all de minimis. That's it was only 5%. And Brandon and I joked that we have about 95% of our problems, too, right? Um and that like mail became especially contentious because data in mail is difficult, right? And we've we've made arguments that, hey, look, mail is mail. And I we fully understand the history and the importance of mail, right? And particularly for the United States and the founding of our country, when our founding fathers were trying to make planning, they were using mail. They were sending letters back and forth. When, you know, our how do you how do you do that today? Well, you still can, but mail has kind of gone into a cargo environment. When they picked up e-commerce and foreign postal operators around the world picked up e-commerce to ship, and that's their money-making business, because okay, stamps are, you know, here in the US they're expensive, and other parts of the world not so expensive, you're not making a whole lot of money on that. But you are on when you're picking up, name your online retailer and you're shipping stuff around the world, that's big money. So when we ask for more data, hey, the data's there. It was just our postal operators weren't equipped to get the data. They are now, because over 85% of the countries are now back to shipping mail back to the United States full force.

SPEAKER_03 13:10

Great example of kind of one of the big risks on the top of your list and what you've had to deal with. All right, so lightning round, because we want to give maybe just a couple of minutes for QA. Lightning round, if you could wave a magic wand and ask for a new tool, a new capability, what would it be? Eric, I'm gonna start with you, then we'll do Shane and Brandon will let you have the last word.

SPEAKER_00 13:36

Yeah, I'll just start off the top with just kind of adding to another risk that you know we haven't talked about was the trans the illegal transshipment risk. And certainly a lot of things that we we are challenged with is the fact that, again, going back to country of origin verification, under making sure that what is represented at time of entry for specific goods is from a specific country that is being declared by the by the importer. So, you know, transshipment is a normal part of uh global trade. You know, if something comes on a ship is made, say in a specific country, goes to another country as a route of transit, maybe gets off of that ship and gets put on another ship and then goes to its ultimate destination. But there's a legal transshipment where potentially that comes off that it gets manufactured in a specific country, goes to in transit to another country, gets stamped as being made in that country. That's something fairly simple. Um the box gets stamped, made in that country, and then gets sent to, say, the United States as an ultimate destination, and then gets represented as being made in country Y has a lower tariff rate, right? And so that's a fairly simple uh evasion scheme. Other things that where we see um where uh parts are being transshipped, maybe or they go to a specific country to be assembled, and then it's being represented as coming from country Y. Um, but the rules of origin saying say that, well, if it's if it's being assembled this country, then it has to be represented as coming from country X. And so in those instances, we have to send folks, potentially send folks on the ground, and we have to look under the hood, maybe under look look within the piece of machinery, and you may find parts that are stamped made in Country X when you know they're bit is being represented as country Y. And so all these things are things that you know, there are methods that you can't really take the human out of the loop, right? I mean, it really requires folks to get on their hands and knees, look under the hood, look, look at the pieces of uh of fabric and the and the equipment to understand the or actual origin. But you know, so if if we're what we're really kind of pursuing right now is looking at the technologies, one where we can leverage technologies again to better understand how trade is flowing and how whether trade is is that as it's being represented to the agency is actually you know truly representative of what's occurring. So I think if I were king for the day and could snap my fingers, really what I would really want to actually get, you know, pursue as an end-state. I knew that I know that, or I believe that you know AI ultimately is going to play a critical role in what we do day in and day out, in the sense that one, it it I think it's functional in the sense that a lot of the processes that we do are still fairly manual or are uh labor intensive, right? So being able to cut down some of that process time to make us more agile in our enforcement. But segmenting wise, what we know with regards to data, like where we see like global shipments and where we see container movements and what we actually see represented on the ground, a lot of that is capabilities they segmented, where we can have a transparent view at certain specific points in time to show that yes, this was here or this is where it was manufactured at a certain point in time, but something that brings it all together, right? Maybe in the future where there could be some predictive uh analysis or some predictive um things that'll help our investigators or analysts. Uh, but then also not only, you know, one of the things that we hadn't mentioned is we do have a significant uh body of auditors, I think is our largest footprint within the Office of Trade, uh, who are looking through and also pouring through as our as our import specialists are doing, I mean, pouring through thousands and thousands of pages of business transaction documents, purchase orders, movement documents, and so something that really, you know, again, shrinks the haystack and shortens the time that that we're able to enforce.

SPEAKER_03 17:07

You're not asking for much.

SPEAKER_00 17:08

Yeah, no.

SPEAKER_01 17:10

I'm gonna go very fast. Supply chain security, visibility, and traceability from supplier to manufacturer to ultimate constantly. If I can get that, I literally can clear trading cargo at the speed of electrons because we we know we can validate and we can make it safe so that when we can truly focus on the threats that the US is facing.

SPEAKER_03 17:32

Excellent. Okay, Brandon?

Closing Thanks

SPEAKER_02 17:35

Yeah, everything both of them said, uh, plus uh, you know, the capability to spot these shell companies, frankly, before they've like run up a bill, right? Frankly, before they've spotted harm. Uh, and then just want to second on the on the supply chain tracing. You know, the ultimate goal here is as we increase operational efficiency. And what I mean by that is if we stop cargo, we need we need to strive to be as like sure as possible that we will find a violation. And so tools that drive that intelligence up.

SPEAKER_03 18:06

And the private sector would welcome not being um the needle in the haystack. And so I think using tools for traceability and uh making sure that they are compliant is is also really important. You've done a great job giving us a view into the trade mission at CVP. I'd like to thank you. Unfortunately, we don't have time for questions, but um we will live to fight another day.

SPEAKER_01 18:32

So thank you. Thank you. Thank you.