MediaTalk | Season 3 | Ep. 20 - An Illinois Lawmaker's Plan to Block Global Chip Smuggling

View Full Transcript

Mike Reynolds: Hi, I'm Mike Reynolds, a senior reporter covering the media industry with S&P Global Market Intelligence's tech, media and telecom news team. Welcome to "MediaTalk," a podcast hosted by S&P Global, wherein the news and research staff explore issues in the ever-evolving media landscape. Today I'm joined by Congressman Bill Foster, a Democrat from the Chicago suburb of Naperville, who has been representing Illinois's 11th Congressional District since 2013. Rep. Foster's district includes Argonne National Laboratory, which has made key advances in quantum science and innovation in artificial intelligence. As the only PhD physicist in Congress, AI, quantum policy and infrastructure are key components of Rep. Foster's agenda. He also serves on the House Financial Services Committee. Before he was elected to Congress, Rep. Foster worked as a high-energy physicist and particle accelerator designer at Fermilab. Joining the Congressman from Capitol Hill today is my TMT News team colleague, Stefan Modrich. Stefan covers technology from DC. Thanks to both of you for joining me today.

Let's start with Rep. Foster. Reports of AI chips made by NVIDIA Corp. being smuggled to China in violation of US export controls have set off alarm bells in DC. Can you break things down for us, the severity of chip smuggling?

Bill Foster: Sure. The problem with high-end AI chips being smuggled into China really started getting highlighted with DeepSeek's involvement in using a vast number of these chips that had apparently been smuggled. DeepSeek, for example, bought about 10,000 chips before there were export controls at all, but then seems to have been training their models on a much larger set of chips that they have obtained in violation of the export controls. This obviously has huge market-moving implications when you see their market reaction to DeepSeek's claims that they did this, frankly, much more cheaply than they actually did. It seems clear that they have at least a $1 billion level investment in hardware that they don't like talking about, frankly, because it's unlikely that most of this was obtained above board. There are also reports that it's actually cheaper to buy these smuggled high-end AI chips and modules, the GPUs, in China than it is in the United States, which gives you an indication of how ineffective our export controls are.

The legislation, which has a companion version in the Senate, is something that would mandate that the manufacturers of high-end chips provide mechanisms inside these chips to detect that they have been smuggled. This involves location verification mechanisms in them. The way this would work is that if, say, you owned a datacenter now full of these chips and you claimed the datacenter was in Singapore, you would put something equivalent to a base station for the Starlink system on top of your datacenter, where the satellites would know within a few feet exactly where that base station is, and it would be impossible to spoof. Then that base station would be connected to the high-speed network inside your datacenter. It would rapidly do a roll call of all the chips you had licensed inside that datacenter, and the individual GPUs would respond to that roll call by responding to a cryptographic challenge that allowed them to prove they are the specific chip manufactured with a specific secret number buried inside it. This would allow you to very rapidly detect that the chips had been smuggled out of the datacenter they were licensed to and potentially into China, Russia, Iran, North Korea, you name it. By mandating that chips have the ability to respond to this roll call, you can make a very inexpensive system. In fact, for most of the chips that NVIDIA is shipping today, they already have the ability to assign a cryptographic challenge with a secret number that's put in at the point of manufacture. This is a really solid and very inexpensive way of verifying that the chips are where they claim they are. Obviously, if you're testing this from time to time and all of a sudden you find that some of the chips are not responding to that roll call, then you can send the police over to find out what's going on.

The other thing is that there's a study that will go on for the next three years with periodic reporting. There are a number of follow-on things that can be done to strengthen this. For example, if you have mandatory licensing where the chips will simply require, before they start operation, to verify that they have a cryptographic license certificate that shows they are licensed for the current period of time, they will simply refuse to operate. That's the nuts and bolts of what we're thinking about in this legislation. We're going to be introducing the legislation, and we're hoping that it should be non-controversial enough that we can maybe even get it on what's called the suspension calendar, which means that as long as you get two-thirds of the vote, you can immediately put it up for a vote.

Reynolds: That would be a nice quick turnaround. Congressman, as Stefan has written about, tariffs have been a source of challenges for the tech supply chain. As I think you told us in Market Intelligence, the AI supply chain is "scattered around our allies in the free world." Given your background as a scientist and technologist, can you share any particular insights from your constituents about how they look to navigate these economic conditions and what Congress can do to alleviate tariff-related burdens placed on businesses at this juncture?

Foster: I think I first designed integrated circuits back in the early 1990s and did it throughout my career at Fermi National Accelerator Lab, or Fermilab. What we were doing there was smashing protons and antiprotons together to make particles that have not been around since the Big Bang. To do that analysis, you needed integrated circuits, computer chips that were beyond the state of the art. Because we couldn't buy them, we had to design and build them ourselves, which I spent a lot of my time doing for most of a decade. I've been just frankly blown away by the way companies from around the world have advanced the state of the art in chip building. Some of the most impressive things to me are actually these machines coming out of ASML Holding NV in the Netherlands, which integrate components from places like OShot Optical in the former East Germany and control systems and high vacuum equipment from around the world into these incredible devices that are simply the key element in fabricating high-end AI chips. These are built in the Netherlands. If you were to make a map of all the places we depend on, it would include big parts of the European Union, United States, certainly Korea and Japan, Taiwan for sure, where the final chip fabrication is done in most cases. All of this comes together, and I guess it's a miracle that we should all be thankful for that these are all in the free world. If you want to build a high-end AI chip, you need to have access to the supply chain that is scattered throughout the free world. For that reason, the free world has a joint responsibility to make sure that these AI chips do not fall into bad people's hands. There's a lot of bad things that you can do with advanced AI, and I think the whole world is aware of that at this point.

Reynolds: Rep. Foster's district includes Argonne National Laboratory. Paul Kearns, who is Argonne's director, told a House subcommittee in February that there are some $37 million in research activities that have been suspended or put on hold amid the government funding freeze. What are you hearing from Argonne in terms of how this has worked and what's going to be the impact here?

Foster: My phone has been burning a hole in my pocket with scientists that I've known for 40 years calling me up and describing how their research activities are being destroyed. For example, there's someone who's very involved in clinical trials for drugs to prevent future pandemics. He has a very widely scattered research enterprise with collaborators from all over the world. Some of that research, things like clinical trials and detailed testing of the chemical purity of the samples, he subcontracts to other countries around the world, like Japan, Korea, and all over Europe. These have now been banned by Health and Human Services. They're just telling you to cancel all these subcontracts in the middle of the period of performance for these grants. This makes it impossible to deliver what you promised for your grant, and it makes it impossible to maintain the collaborations that often have taken more than a decade to build. The damage to the scientific output of our country and the free world is enormous. Then, in regards to the national labs as well as laboratories at universities, these are enterprises that take decades to build, and with three months of zero funding, you can destroy something that's taken 30 or 50 years to build. The people that have come together to do their work there are often people who could have a variety of jobs. They're tremendously talented people and could get a new job at a drug company, a different university, or a different country by snapping their fingers. When you tell them, "I'm sorry, you can't order any of the equipment you need to do your research, or you can't hire any students to come and help you with your research," it's really discouraging to the young students who are thinking of getting into the field. It's very discouraging, and people say, "I think I'll do something else with my life." That's going to be the biggest long-term damage from what's being done.

Stefan Modrich: You've introduced a bill with your Illinois colleague, Sen. Dick Durbin, the American Innovation Act, which proposed to grow the budget of the National Science Foundation 5% annually, plus adjustments for inflation. This comes as you've alluded to, that the Trump administration has proposed to cut the NSF's 2026 budget by 56% to $3.9 billion. So where will the impact of these budget cuts most likely be felt throughout the tech sector?

Foster: They'll be felt immediately in the ability of the tech sector to hire the kind of talent they want. If people don't go into the scientific fields that really feed the whole AI train. If you look at the history of AI, there was what's called the AI winter, a period of probably a couple of decades where no one was interested in AI. Who was carrying the torch and keeping the flame alive for AI was the federal government, who was funding mathematicians, computer scientists, philosophers, and neuroscientists to keep trying to push the boundaries of what was understandable. Then all of a sudden, they had a breakthrough, and AI became what we see today. Now it is large enough and successful enough that you can see companies coming in and putting in more capital, frankly, than the federal government ever will be able to. But what you lose when you cut the National Science Foundation budget by a factor of two are all of the researchers who are working on the next big thing, the thing that we don't think is important now but might be, and that will someday soon bloom into something of incredible importance. You see that over and over again, where small research areas turn out to be a lot more important than even the researchers working on them expected.

Modrich: You also reintroduced a bipartisan bill, the Keep STEM Talent Act, to make certain advanced degree holders eligible for permanent residency status. We've reported a lot about this; the Semiconductor Industry Association has an estimate projecting roughly 60,000 to 70,000 of the potential 115,000 new high-skill jobs created by the industry risk going unfilled by 2030. How would the bill that you've drafted help retain that top tech talent? How does it fit into this overall macro picture of the explosive growth that we've seen forecasted for AI and advanced chip production in the years to come?

Foster: It's something that you will hear from every tech executive, that they cannot find the talent they need. Much of the talent they need is actually trained at US universities, but then we have this unfortunate policy of pushing them back to their home countries when they've completed their PhD degrees or their engineering degrees or whatever education they've received, and this is actively stupid. Even President Trump would regularly complain on the campaign trail, "Why is it that we educate these people, give them PhDs, and then send them out of the country?" This is the reason we're reintroducing something that we've been trying for probably almost 10 years now. We're reintroducing it with a Republican partner, the Keep STEM Talent Act, which would simply, roughly speaking, staple a green card to a PhD degree for STEM graduates.

Reynolds: At the end of 2024, Congressman, the House released its bipartisan task force report on AI. I believe it included some recommendations on secure identity infrastructure. Can you tell us why that's an important area of focus? Is there a deepfake of you out there maybe making physics jokes? We want to know, have you heard the one about how you can't trust an atom... because they make up everything?

Foster: I think several of my physicist former collaborators and colleagues have made various deepfake jokes. As soon as the technology became available several years ago, they decided it would be amusing to make a deepfake of the physicist Congressman Bill Foster. One of them I remember was repudiating our discovery of the top quark when we were smashing protons and antiprotons together to make particles that haven't been around since the Big Bang, and we were able to produce what's called the top quark. It's a single subatomic particle that weighs more than an entire atom of gold, and for various theoretical reasons, it's quite possibly going to be the heaviest form of matter ever discovered. This was a great discovery at Fermi National Lab, one that will be in the science textbooks forever. But they thought it would be amusing if they made a deepfake of me announcing that it was all a hoax. It was quite amusing at the time, but it's become a lot less amusing since that deepfake technology is now being used to scam thousands of people, and with advanced AI, it's going to be able to scam them at scale. Deepfake identity fraud is a huge and growing problem, and it's even hit banks.

I'm the lead Democrat on the banking subcommittee on House Financial Services, and there was a really alarming heist of $25 million where the chief financial officer of a financial company was invited to a Zoom with his colleagues and his boss. They had a long discussion of a project they might be interested in starting up. They concluded, yes, we should start up the project, and the CEO ordered the financial officer to get the project started by transferring $25 million into a specific account. The CFO did not realize that everyone on that Zoom was a deepfake impersonation of his boss, his coworkers, and so on. That's the level of sophistication we're getting to. So what can Congress actually do about this? I think it's unlikely that we're going to be able to have software that will be able to really detect the difference between a deepfake and the real image. We're going to try, though. We will make some difference, but ultimately, I think it's going to be a losing game. What we can do is provide people who want the ability to prove they are who they say they are and not a deepfake impersonation of themselves. This is the other side of that coin. For example, if someone wants to send me a text message or an email, unless they're willing to prove that they are not a deepfake impersonation, that they are actually the person they claim they are, which you can do with a secure digital ID, this is, I think, the best thing that Congress can do.

We actually have the tool at hand for that, which is the digital driver's license or the mobile ID. These are being rolled out in states around the country. Now that most everyone in the United States has a Real ID compliant driver's license, it's a small additional step to put that on your cell phone. That allows you to combine your government-issued Real ID, which proves you're a specific legal person, with your cell phone, which can prove it is the unique device in the world with cryptographic techniques. Then take advantage of the fact that a modern cell phone can recognize its owner with biometric login. You combine those three together, and that's the best that you can do to know you're not being scammed by someone who is just operating anonymously on the internet. The tool at hand is the mobile driver's license, the digital driver's license, and by getting these federally recognized, we can prevent billions of dollars of fraud. There was a huge problem with COVID identity fraud, where scammers absconded with tens of billions of taxpayer dollars with fraudulent ID because we did not have such a system. Countries that have done this, like Estonia, like the entire EU, by the end of next year everyone in the EU will have a digital ID that they can represent if they wish to use one, and so this will be a game changer for fraud prevention both for the government and taxpayers and for industry, which is very excited about being able to use this.

Modrich: So Congressman, Illinois native here. I gotta ask you this is gonna be a bit of a curveball, but I'm a Gene & Jude's guy. What's your favorite hotdog in your district?

Foster: That's a tough one because I represent pieces of eight different counties in the Chicago metropolitan area with lots of really good hotdog options. But I'd have to go with a Mayo Grill in Aurora. I've now made enemies in seven of the eight counties of my district.

Reynolds: You're gonna get in trouble. That takes us to the end of this episode of "MediaTalk." I just wanted to thank Congressman Foster and my colleague Stefan Modrich for taking a lot of time to talk about the ever-expanding world of AI policy. Thanks, fellas, for taking the time here. We appreciate it.

Foster: Yep, I'm happy to be here.

Modrich: Thank you, Mike.

Reynolds: Thank you, guys. Thanks to all of you for listening. We'll catch up soon on the next edition of "MediaTalk."