Reflecting on the Commerce Department’s Role in Protecting Critical Technology with Under Secretary of Commerce Alan Estevez

Photo: CSIS
Available Downloads
This transcript is from a CSIS event hosted on January 14, 2024. Watch the full video here.
Navin Girishankar: Good afternoon. I’m Navin Girishankar, the president of the Economic Security and Technology Department at CSIS. Welcome to today’s event. I am delighted to host Alan Estevez, Undersecretary of Commerce at the Bureau of Industry and Security. Former colleague. I’ve had the pleasure of working with Alan over the last three years. And just absolutely delighted that he’ll be joining us here today.
And obviously you may have heard that export controls is in the news recently. I think more so today and maybe even tomorrow, I don’t know. But Alan has really overseen BIS during a period of a meaningful and significant evolution of export controls, as many of you know. And over the years, seen him work tirelessly with his team, oftentimes below the radar screen, working hard to ensure security of U.S. markets, products, sensitive technologies. And so really want to salute Alan and his team before they come up here.
Probably as he’s kept us busy at tracking what the administration has been doing, no one has been more focused on it and busier than Greg Allen, who is the director of our Wadhwani AI Center. And so I think no one better to have this conversation with Alan than Greg. And so with that, let me ask Alan to come up and really just thank him for making time available today. (Applause.)
Alan Estevez: OK, thank you for that very nice introduction, Navin.
You know, when I used to run logistics for the Department of Defense, and I would talk about supply chain, people used to, like, kind of go into this kind of glaze. Like, he’s talking about supply chain again? I don’t want to hear about supply chain. Fast forward, you know, during COVID, everyone wanted to talk about supply chain. And they still do. I don’t know what it was like when you were – had my job, Eric, or when – Bill Reinsch is somewhere in here – had my job. But, you know, I used to kind of glaze over when people talked about export controls too. (Laughter.) And I don’t think that’s the case anymore. At least we’re trying not to make it the case. And, you know, we’ve had a little bit of the cadence over the last couple of weeks of – I think this week it’s a rule or two a day related to some important things around artificial intelligence and our ability to protect the nation against our adversaries.
So let me talk very briefly about a couple of things that I think we’ve accomplished in the last four years of the Biden-Harris administration – my three – almost three years in this seat leading BIS, which it has been a great honor for me to do. And then, Greg, you and I can have a wonderful chat up here about anything you want to talk about.
But if you look back over what we’ve accomplished, you know, most of the controls we’ve put on – and I’ll talk about three things, really – are controls related to the PRC or controls related to Russia. Obviously, the regular business goes on related to nuclear programs around the world or chem-bio programs around the world and those type of things. So that business has been in a pretty steady state. But our business around the PRC has gotten a lot of notice; our business around Russia has gotten a lot of notice. And we stood up a brand new office called the Office of Information Communication Technology Services, ICTS, that is also making a little bit of a splash these days. So let me talk about those three things, and again, then we’ll just jump into some Q&A because I think dialogue is way more important.
With regard to the PRC, you know, people always ask, like, what are these controls related to China? And our controls really impact the highest end of tech. So there’s lots of trade going on, but the controls that we have put on – our October 22 controls related to semiconductors; October – November 23 – I don’t know, it might have been October 23; December of last year related to semiconductors have all been about the highest end of tech, and that highest end of tech really related to artificial intelligence because – and then our rule yesterday on AI diffusion related to artificial intelligence. Because if you think about artificial intelligence from a military perspective, artificial intelligence has lots of uses for commercial applications. Artificial intelligence has some game-changing capabilities that can help all of us in our daily lives going into the future.
It also has some malicious applications to it. You know, most people think about the deep fakes and, you know, news-related issues around artificial intelligence. But artificial intelligence for, like, military application enhances command and control, enhances targeting, enhances logistics, enhances autonomous warfare. All those things are very worrisome. On top of that, artificial intelligence at the next generations of models – not the models that are there today – are going to facilitate cyber capabilities – cyber warfare capabilities. It’s going to facilitate biological development capabilities. So all sorts of things that artificial intelligence can be used for, for purposes that go against the national security interests of the United States and its allies.
So the controls we put on semiconductors and semiconductor equipment going to the PRC have all been about impeding the PRC’s ability to build the large-language models that can threaten the United States and its allies from a national security perspective.
Now, I’m a realistic person. Will we stop the PRC from developing models? The answer there is, you know, no. The realistic answer is no. Over time the PRC will – they have very smart people, very good engineers; many of them went to the same universities that our top engineers went to, and they’re going to work around, develop new methods and new techniques and new technologies. But our point is to not them let our state-of-the-art technology to advance their capabilities, and that’s what our focus has been there. And I would argue that no administration has been more focused and tougher on the PRC with relationship to that.
And I will give credit to the previous Trump administration for starting some of the things that we took on that path. But we really took it to another level – industrialized it, if you would – issuing country-wide controls versus specific entity-focused controls. Growing the allied base around those controls have been really critical and I think have impeded the PRC’s ability to develop the highest-end chips and to develop those AI models that will threaten us in the near term.
All that said, the United States still needs to run faster, right. We still need to be focused on the innovation base of the United States. Anything that we’re doing in the export-control side is a hurdle, a bump in the path. It’s not going to be a dead stop. We need to keep out-innovating in order to remain ahead of the PRC on that.
With regard to Russia and Russia’s further invasion into Ukraine starting in 2022, you know, we always had some significant controls on Russia, but the team at BIS – you know, most of this started before I got there in April of 2022 – build a coalition of 38 nations that put significant controls on the Russian industrial base and on exports going to Russia. And we tightened those controls in conjunction with our allies around the world repeatedly. I think we have 50-plus rules, you know, multiple entity listings – I’m looking here, like, a thousand Russian entities on the entity list, 500 since the invasion, related to Russia’s ability.
Again, you know, Russia has worked around some of those controls. If you want to talk about the key component of working around those controls, you have to go back to talk about China and China’s facilitation of the Russian industrial base. Nonetheless, we have imposed costs, made it harder. That’s why you see Russia going to North Korea for weapons and soldiers, why you see Russia going to Iran for weapons and building a kind of true axis of evil, if you would, to work around.
But that doesn’t make our controls not successful. Imposing those costs and slowing down and impeding the Russian industrial base is critical to the national security of the United States, and certainly to the people of Ukraine.
Finally, on the ICTS, you know, I got to the BIS, and ICTS was about four or five people, all borrowed manpower, sitting in an office with no money, no funding; a directive to stand up this office but no money, no funding. We finally got funding in January-February of ’23. Since that time we have hired an extremely accomplished director for that office, Liz Cannon, who’s a career official, and she has built an office of about 80-plus people right now.
In the last year we put a ban on software related to Kaspersky Labs. We all know what that is. That was a virus software that is embedded on people’s laptops and then their business systems. And if you want to talk about cyber risk, talk about a software that has access to the hard operating drive of your computer and think about the risks related to that when it’s an adversarial nation-controlled piece of software.
Just today we finalized a rule related to components, key components of automobiles from the PRC or from Russia and then full-up cars that contain those components. We started off today – this rule is about passenger vehicles. We have in process some work around commercial vehicles that will build on that. Supply chains are a little different from that.
And why are we worried about cars from the PRC? If you own a car, a connected vehicle, a fairly new car – let’s say 2016 forward – and your car gets a software update, which is probably most of the people in this room have a connected vehicle – your car knows a hell of a lot about you. Your car knows probably more about you than your spouse or your friends know, because your car knows where you go all the time, as long as you’re in your car, right? That little GPS device knows where it’s going. And it’s keeping track of that, by the way. Your car probably has your phonebook in it. And your car knows who you call and, frankly, can record those calls. All that software could come in means software can go out. So there’s risk of data. There’s risk of cyber intrusion.
There’s risk that the software update to your car can say – what’s today, January 14th? The software update to your car could say, hey, on January 20th when you push that little button on your car that says start – we used to have keys. Some of us do still. The car does not start. The car becomes brick.
That’s a risk to the critical infrastructure of the United States so that’s why we are banning key components of vehicles that run the automated systems and run the software systems of automobiles that come from the PRC or related things like that.
We’ve kicked off something on drones related to the PRC and we have a number of other investigations ongoing. So that office is full up cranking. Really proud of standing that up.
The final thing I’ll note, you know, I do have an enforcement arm, and it’s not the final thing. I’ll have one thing after that. You know, enforcement is a tough business especially when enforcement used to be stop this big item from leaving the United States. Now most of the stuff that we’re protecting, frankly, a lot of it isn’t even made in the United States. It’s being covered either through allied agreements or it’s covered under something called foreign direct product rule.
That’s a much tougher thing, and a lot of it is things like semiconductors which some of the semiconductors we’re talking about are actually pretty big units. But, still, it’s much harder to control than a large CNC machine, for example.
So, you know, we stood up, along with the Department of Justice, a disruptive technology task force that has put a number of people in jail including someone got sentenced yesterday related to Russia’s support. We’ve done a number of big administrative settlements. But tough job there.
So the last thing I’ll say is BIS resources. BIS – we’ve done all this under a resourcing scheme that’s essentially been the same since 2010. My budget has essentially been flat aside from the bump up I received for the ICTS program since 2010. Received a little bit of a bump up during export control reform during Under Secretary Hirschhorn’s time. But other than that, pretty flat.
Meanwhile, you know, I don’t know if any of you look at the rules that we put out other than the headlines but they’re pretty complex damn rules, right? They’re not like 30-page rules anymore; they’re 250-page rules – if you remember the export bar, like, on making big houses for you – and they’re complex, and the licensing has doubled or more since that time because I’m controlling a lot more stuff and those licenses have become more complex.
Meanwhile, the number of people, flat. And I have an IT system that came into place even before Eric was the –
Audience Member: I had Lincoln’s computer on my desk.
Mr. Estevez: Yeah. There you go. Yeah, I’m working with McKinley’s. (Laughter.) That is not the way to run a business. I can’t say, hey, Siri, what listings did we give to an Entity Listed party either before or after? That is a manual search in my system.
So I need a new IT backbone. I need more gumshoe, as far as agents. I need more licensing officers. And I need applications – I’m going to say the word Palantir – but things like Palantir to help my agents do tracking. I need supply chain illumination tools and all sorts of other apps. And, frankly, I could use artificial intelligence in this space, too.
So it’s a wide-open space. BIS needs more resources. And I know Greg’s a big proponent of that, too, so I’m teeing you up for a question later.
And with that, Greg, let’s sit down and talk. (Applause.)
Gregory C. Allen: If you could sit on this side. Thank you.
Well, Mr. Undersecretary, thank you so much for those fabulous remarks and thank you so much for coming back to CSIS to talk in just the last couple weeks of the Biden administration, which is really not a sleepy couple of weeks in your case. You had, as you said, a rule come out yesterday, a rule come out today. I think –
Mr. Estevez: Two rules tomorrow.
Mr. Allen: Two rules tomorrow. So you’re speeding up, you’re not slowing down, across the finish line.
Mr. Estevez: Plus a massive rule at the beginning of December.
Mr. Allen: Yeah. That was no small rule, I should say.
Mr. Estevez: And I would note that, you know, whatever was going on in the world tried to make it harder, two snow days, the presidential funeral.
Mr. Allen: Yes.
Mr. Estevez: I was expecting locusts or frogs or something. (Laughter.)
Mr. Allen: And yet, you got it done. And here we are. So I want to start, if it’s OK, with you. Because I learned something about you that I didn’t know previously, which is you were not asked to take this job. You were told you were going to take this job. (Laughter.) And by one of the great luminaries of U.S. national security, the late great Ash Carter. And so I wonder if you could just tell a little bit of a story about, as you took this job, what was on your mind?
Mr. Estevez: Sure. So the way that came about was, frankly, Secretary Raimondo called me, cold called me. It was during COVID, so it was a Zoom call. She and I had about a 45-minute conversation, at the end of which she said, hey, you’re the guy. I want you to take this job. You’re ready to come on board. You know how to do this stuff. And I said, you know, secretary, I’m really comfortable here in the private sector. (Laughter.) I’ve done my time. I’ve done 36 years at DOD. I’ve served the nation. I feel good about myself. I’m good. You know, I’ll give you some ideas. And she said, yeah, that’s the wrong answer. We’re staying on this call until you say yes. (Laughter.)
And I did say yes by the end of that call. And then the next day, Ash Carter, one of my – you know, a great friend, God rest his soul, great mentor to me, former Secretary of Defense, and I worked for him in a number of other jobs, called me and said: Hey, Alan. Gina Raimondo called me. She wants you to take that BIS job. You’re taking that job. Click. (Laughter.)
Mr. Allen: Which is very classic Dr. Carter. But I think it’s worth pointing out, and this is something that Bill Reinsch, my colleague here at CSIS, has pointed out, is – and we’re in a presidential transition moment here right now. So there’s the list of jobs, you know, that the new president is responsible for appointing. That list of jobs is called the Plum Book. But there’s a less well-known list of jobs, which is called the Prune Book, which are the jobs that are really important and no fun at all to have. And BIS undersecretary is under that job. (Laughter.) And I’m glad to see you crack a smile that you maintain, you know, a good demeanor as well.
Mr. Estevez: Throw some water on it, it becomes a plum. (Laughter.)
Mr. Allen: Yeah. But really, one of the hardest jobs in government, I think one of the hardest times to have one of the hardest jobs in government. And so, on behalf of all of CSIS, I want to thank you for your years of public service in this job and your many prior jobs.
I want to now start by taking us back to October 2022. This was when the October 7th, 2022, export controls came out on artificial intelligence and semiconductors. I thought it was a landmark moment in the history of international relations. Really, I think probably the second-most important thing in foreign policy that happened that year, other than Russia’s invasion of Ukraine. And I want to take us to a statement by Secretary of State Antony Blinken, who said, “We are at an inflection point. The post-Cold War world has come to an end and there is an intense competition underway to shape what comes next. And at the heart of that competition is technology.”
And this was only 10 days after the export controls policy came out. And he really seemed to say that with this new export control policy we are sort of bookending the end of the post-Cold War era, and this new policy is sort of the starting point for what our approach is going to be writ large. And so I want to ask, you know, do you see your own work in those same historic terms? And, of course, this has a – this has a resourcing follow up question, but let’s just start there, your reaction to that.
Mr. Estevez: Yeah, look, I think – I’m a modest person, but I think the work that BIS has done is critically important at this time. This is not, you know, to use a commercial analogy here, your grandfather’s BIS, right?
Mr. Allen: Yeah. Maybe the computers, but not the work.
Mr. Estevez: (Laughs.) Exactly right. We are certainly hot, dead center in national security strategy. The amount of time I’ve spent in the Situation Room is way more than I expected in this job. And that’s because technology is critically important in this space. So, you know, look, when I’m sitting on a Sunday afternoon on a phone call with the Secretary of Commerce and the National Security Advisor talking about the minutia of the parameter of some export control rule, you say, huh, this is probably really important stuff. Shouldn’t Jake be doing something else? But it is critically important. And I think we’ve risen to meet that moment.
Mr. Allen: Yeah. I certainly agree, and I think – now, that policy, in addition to making new big houses for the lawyers who service this work, as you mentioned in your remarks, was, you know, followed on. And so I’d like to – sticking with the AI and semiconductor story here, if you could just sort of explain your own looking back sense of what was the big moment that – what happened in October 2022, what happened in October 2023, what happened in December 2024, and what’s happened in January –just yesterday, in fact, with this diffusion rule.
Mr. Estevez: Yeah.
Mr. Allen: What do you see as the stairstep in your own thinking and in U.S. policy?
Mr. Estevez: Sure. So let me start off with what’s consistent across all those, and it goes back to what I was saying from the podium, that we were focused on the risks related to artificial intelligence – the national security risk related to artificial intelligence and the need to put some control on that and certainly to control adversarial use of that against us.
And again, you know, in the case of the PRC, in the case of any nation that we have controls on, they’re sovereign nations. PRC can modernize their military; they just shouldn’t be doing it with our stuff.
Mr. Allen: (Laughs.) Yeah.
Mr. Estevez: And so that’s point one. So you have a threat vector here, and you know, consistency of what’s across that threat vector. And you know, I’ll throw in the small yard-high fence thing and what does that mean, because people are going to always ask me, well, what’s the definition of the yard? Well, the yard is really defined by the threat and the technology. And technology moves, right? So when we put out that October ’22 rule was before the explosion of ChatGPT –
Mr. Allen: Yeah.
Mr. Estevez: – when everyone said, oh, this is a real thing, not some like “woo-woo,” you know, like, deep inside JAIC or where you came from.
Mr. Allen Yeah, yeah.
Mr. Estevez: – which needs to do more, too, by the way. (Laughs.)
Mr. Allen: We had some fun stuff but we did not have ChatGPT.
Mr. Estevez: And so, you know, we put out that rule related to protection of the most advanced semiconductors that were on the market at that time and the tools to make those semiconductors. Well, two things occur in between there. One, we didn’t get the parameter exactly right.
Mr. Allen: Yep.
Mr. Estevez: You know, one of the things I noticed when I came into this job is that I’ve never made a semiconductor, and frankly no one on my team had ever made a semiconductor. I shouldn’t say no one; there’s probably someone. But, you know, suddenly I had this CHIPS office where I had people who actually did make semiconductors.
Mr. Allen: Yes.
Mr. Estevez: So you start talking to them, you start talking to the companies. You know, people say we’re too close to industry talking to the companies – in order to understand, like, what makes a good artificial intelligence GPU, I spend a lot of time with people who either built you know, the model – big, large language models – you know, people at OpenAI or Anthropic or Inflection – you know, name your AI company du jour – or I talk to Nvidia and AMD and Intel and the people who make chips. So to, like Samsung, you know, how do you make a good chip and what goes into that?
So we realized we had the parameter wrong. So we adjusted the parameter – I think that parameter came out much closer to like –
Mr. Allen: Yeah, this is moving from processor speed and interconnect speed to total processing power. Yeah.
Mr. Estevez: That’s correct.
Mr. Allen: Yep.
Mr. Estevez: And so that was the October 23 rule.
All right. The PRC is not stagnant. They’re saying, OK, I can’t get an EUV machine; I can’t get the highest DUV machine because I have countrywide bans, and you know, there’s controls on certain fabs. Somewhere I’m going to stand up fabs and I’m going to start doing some techniques that may not be economically viable – multi-patterning –
Mr. Allen: Yep.
Mr. Estevez: – that TSMC had tried in the 2010s and then waited for EUV machines before they went down to that level – that, you know, if you were going to do it from an economic standpoint, you’d fall on your face; but if you’re subsidized and the economy of scale isn’t your worry – I can, like, produce chips. I can’t produce high yields but I can produce a lot of chips at low yields.
We said, OK, what do we need to do about that? So, you know, again, the adversary has a vote, just like the enemy has a vote on a battlefield. And we made adjustments, and those adjustments were reflected in the December 2 rule of this year.
Mr. Allen: Of last year.
Mr. Estevez: Yeah, of last year, of last year.
The AI diffusion rule that we put out yesterday is again about, you know, the tech ecosystem around artificial intelligence and the data centers and how those data centers are being used and how do you protect model weights around the world, because model weights can be stolen, one; two, people can access models and then do their inference back in their own country around those models. They’re probably not going to do any training. You know, if some adversarial country wanted to do nuclear modeling on a(n) open-source model, like, have at it. I mean, I think we would probably want to look at that.
Mr. Allen: Yeah.
Mr. Estevez: But what we did in the diffusion rule is I think – it was some hard work. We set it so that, you know, normal business can go on. Almost anywhere in the world you can access a large number of chips, some with the license capability, some through VEUs, some through government-to-government agreements, and some through working with U.S. companies. There’s only a few companies that hyperscale around the globe anyway.
Mr. Allen: Right.
Mr. Estevez: And so we set it. And then, you know, if you’re buying low volumes of chips, like you’re a bank building your server farm for your own calculations, that’s not going to register. That doesn’t even require a license.
Mr. Allen: Yeah.
Mr. Estevez: Seventeen hundred the cap there. And so, again, it’s protecting against the nefarious uses of artificial intelligence that our adversaries would try, to do some security requirements around data centers around the world that go into the approval of a VEU-type thing, and allowing commercial trade to go on.
Mr. Allen: Yeah. So I want to – I think that’s an excellent summary of sort of the action process and the learning process of the Biden administration across AI and semiconductor export controls. I want to just talk a little bit about, you know, what you see as the impact of these controls. And I’ll give a few datapoints that I think are especially salient.
So, number one, the Chinese AI firm DeepSeek, which is usually regarded as the best frontier AI model developer of China, at least at the present moment, they released an open-source model that is, in some performance parameters, really competitive, you know, with what’s coming out of Meta or what’s coming out with everything else.
But I think one of the really important datapoints there is that this model was trained on the H-800s, so exactly as you said, you know, getting the performance threshold for the chip restrictions wrong the first time around.
And what does the future look like for a company like DeepSeek? The CEO of DeepSeek, in a recent interview, said the number one challenge facing his company is not financing. It’s not human resources. It is those export controls. That is the biggest challenge facing the future of his company, which I thought was really interesting.
And then there’s the question about, you know, not just buying chips but making chips locally in China. I had the opportunity to speak to somebody who was, you know, talking to folks in Huawei’s supply chain in the very recent past. And what they said is that SMIC, Huawei’s preferred logic chip manufacturer for AI chips, is still stuck making fewer than 20,000 wafers per month.
And the bottleneck in their process is exactly as you predicted, deposition and etching equipment that is on U.S. export controls, which I think is a phenomenal outcome. And it’s not just that they’re bottlenecked; they can’t scale up production in terms of wafers per month. But the yields are dreadful. They’re stuck at, as of November 2024, 20 percent of the chips that come off that line are actually usable.
And so I think, as a direct result of these export controls that we’ve put in place today, you know, the alternative to American AI chips is not Chinese AI chips. The alternative to American AI chips is no AI chips. And I think those are really strong datapoints as an endorsement of the actions that you’ve taken.
I just want to ask whether or not you agree and whether or not there’s anything else that’s salient in your mind as you think about scoring your own homework.
Mr. Estevez: I do agree. But again, I’m always hesitant to say we’re doing well –
Mr. Allen: Yeah.
Mr. Estevez: – because that’s the road to oblivion
Mr. Allen: Yeah. (Laughs.) Only the paranoid survive, as the chip industry often says.
Mr. Estevez: If you’re taking my job, you should be a paranoid schizophrenic for sure.
Mr. Allen: Yeah. (Laughs.)
Mr. Estevez: But anyone who works in Washington, as you know, has to, like, live in the paranoid, at least in the national-security space.
Mr. Allen: Absolutely.
Mr. Estevez: If you’re not living in a paranoid bubble, then you’re in the wrong business.
Mr. Allen: Yeah.
Mr. Estevez: You know, as I was talking about cars – no one should get into their car, right – (laughs) – showed.
So I think we’re doing well. But I think you can’t rest on that; you have to be vigilant. Give them a lot of credit for, as I said earlier, you know, smart engineers, smart capability, and they’re pouring a lot of money into this. So where the next breakthrough can come from, it could come from there; like, here’s the new machine that we weren’t expecting.
Mr. Allen: Right.
Mr. Estevez: No one wants to see a black swan.
Mr. Allen: Yeah, there’s no time to take a victory lap.
Mr. Estevez: So that gets back to the, you know, point I made, and I think Secretary Raimondo made it in one of her closing interviews, is that export controls in and of itself is not the answer to this security risk.
Mr. Allen: Necessary, but not sufficient.
Mr. Estevez: Right. Absolutely necessary things we need to do, and we should do, and I would advise my successors to continue doing those type of things. But we need to be thinking about our own innovation ecosystem and how do we expand that. And we need to think about, you know, from a DOD perspective, how do we start, you know, jumpstarting – I know, like, there’s lots – a zillion articles around this. And we could talk ad nauseam about how do you infuse DOD power around AI and –
Mr. Allen: Big news came out of that today. Oh, sorry, you didn’t mean the electricity part of it. You meant the war fighting part of it. My old life. So I want to ask about, you know, some of the criticisms of – that you’ve suffered, that BIS has suffered, that this entire strategy has suffered.
Mr. Estevez: I don’t read my criticals. (Laughter.)
Mr. Allen: But I just meant the idea that these export controls are accelerating China’s indigenization efforts, that they are strengthening the incentives to de-Americanize. And I want to ask, you know, how you react to that criticism and how it factors into your strategy.
Mr. Estevez: So I believe President Xi. I think, you know, when they put out 2015 China 2025, or whatever timeframe that was – it was around then.
Mr. Allen: Yeah, made in China 2025, yeah.
Mr. Estevez: You know, unlike here, right, central managed, constructed with weird prohibitions in that mix, they’re out doing what they want to do, right? So they’re spending a lot of money on it. I’d rather them spend money on trying to build a semiconductor sector than building a seeker and a missile. But – (laughs) – they’re pressing forward to do what they want to do. So –
Mr. Allen: And they were doing that before the export controls.
Mr. Estevez: That’s right. That goes back in time, and not just in semiconductor space, right? In other spaces too.
Mr. Allen: Right, you mentioned – you mentioned EVs.
Mr. Estevez: Look at cars. Look at –
Mr. Allen: We didn’t have any export controls on EVs and they still managed to do what they can to –
Mr. Estevez: And it’s not just EVs there. It’s any connected vehicle. My internal combustion engine car takes a software update that can make it a brick.
Mr. Allen: Right.
Mr. Estevez: So our belief is that their drive to indigenization has nothing to do with export controls. Does this irk them and drive them to, like, you know, recognize again, oh, yes, it’s lucky we’re doing this? Probably. But, you know, the readings that I read – and I’m reading a lot of readings in other rooms – indicate to us that that was the path they’re on. There’s really nothing that we’re doing that’s expediting that path.
Mr. Allen: Right. And in fact, many of the things you’re doing are making it harder, right?
Mr. Estevez: Yes, exactly right, including putting 120 Chinese indigenous toolmakers on the entity list and denying them the components they need to replicate the tools that they’re reverse engineering. (Laughs.)
Mr. Allen: So I think, you know, as you said, that the resources that China is throwing at this problem are really staggering, right? Literally in the tens of billions of dollars annually for various parts of this equation. And they are very committed to coming up with their own technology, to de-Americanizing. And they’re also quite committed to smuggling in these AI chips. And what – you know, I told you when you were here at CSIS back in, gosh, I guess December 2022, you know, it feels like the budget for Russian smugglers has 10X-ed, and the budget for Chinese smugglers has 10X-ed, and your budget has remained flat. And it just strikes me as unacceptable.
And one of the things that you said on the podium is, I need more resources. I need more resources. Which maybe sounds – makes it sound more self-interested. But you’re not going to be here in two weeks. It’s really your successor, you know, who you’re trying to advocate on behalf of.
Mr. Estevez: Well, absolutely.
Mr. Allen: Yeah.
Mr. Estevez: The institution needs more resources. I want my successor to be successful in that job. You know, there’s, frankly, bipartisan support for more resources. How you do that is the trick on the Hill. But we need more resources. And frankly, even the IC needs resources that are more focused on this sector than on traditional IC assessment areas.
Mr. Allen: And this is – this is one –
Mr. Estevez: And they’ll be the first people to say it. I mean, I’ve talked to David Cohen, talking to Tim Howell (sp) last night about this very topic.
Mr. Allen: Yeah. I’ve read some of the declassified CIA reports of the Cold War. And they did a lot to support enforcement of export controls. They did a lot to support enforcement of semiconductor-related export controls against the Soviet Union. And I’m, like, dust off the playbook, guys. You need to be really good at this again. You know, the BIS should be one of your top customers.
Fabulous. So in just a moment, we’re going to take questions both online and from folks in the audience. So my colleagues will be going around with a microphone. So if you want to signal your intent to ask a question, we’ll do that. But let’s start with some questions that we got online because those are already ready to go.
This one is kind of spicy, so – and you’ve already kind of addressed this to a certain extent, but I want to ask it –
Mr. Estevez: Is the answer yes or no? (Laughter.)
Mr. Allen: Well, so the – this comes from Christine McIntyre (sp), who didn’t identify her affiliation: In a recent interview with the Wall Street Journal, Secretary of Commerce Gina Raimondo stated, “Trying to hold back China is a fool’s errand.” It appears to be in reference to semiconductor export controls. Is she calling the BIS strategy foolish? How do you respond to that?
Mr. Estevez: You know, I’ve already, like, said multiple times here we are hurdles in this space. Sometimes I feel like I’m running down an alley throwing garbage cans behind me, and unfortunately, I’ve been training to run. So, like, you know – (laughter) – because you got to run faster.
Her point in that article – and, you know, there’s a lot more context around what she said in that article – was that the money that we’re pouring into chips and into our own indigenization of chip capability for national security purposes in the United States is critical to advancing national security, not that what we’re doing in BIS is worthless. There’s no stronger advocate for resourcing BIS than Gina Raimondo.
Mr. Allen: Got it. So this gets back to the necessary but not sufficient, not on point.
Mr. Estevez: That’s exactly right.
Mr. Allen: OK. This comes from – OK, another spicy question. This comes from Peter L. Often former BIS officials become lawyers or lobbyists for companies who are advocating for weaker export controls. When is this or isn’t this ethical?
Mr. Estevez: You know, I live in Washington – (laughs) –
Mr. Allen: Yes.
Mr. Estevez: – all right, where businesses talk to the Hill who then talks to me. That’s part – that’s what businesses do. Businesses are in the business to earn a living, to make money, right? And we want an innovative sector in the United States that – I want Nvidia to be successful. I want applied –
Mr. Allen: Right. We want American companies to succeed.
Mr. Estevez: You know, I think that’s critically important for us in this ecosystem. So I have no problem with companies going and, like, not liking exactly what I’m doing. I talk to those companies all the time. They know where I stand. You know, I can’t say that – look, lawyers earn a living from clients. That’s fine, too. People deserve to have the best representation.
We make decisions based on the national security issues in front of us. You know, to me, 36 years at DOD – I think that I was quoted as saying this in a New York Times article – plus this job, national security is my North Star. I don’t know how to do it any different.
So I think that doing this is going to be critical and happens to impact the company in some way, you know, I have to make that choice. Now, if I think, huh, there’s a trade space here that I don’t want to impact the company so much that I lose the innovation space –
Mr. Allen: Right. Right.
Mr. Estevez: – then that’s a national security risk, too. So, you know, walking that tightrope trying to figure out that balance that’s what makes it a prune job. (Laughter.)
Mr. Allen: Yeah. Right. Right. Exactly.
Mr. Estevez: But you have to. You know, companies talking that’s their job. It’s not unethical at all.
Mr. Allen: Got it.
This comes from Demetri Sevastopulo of the Financial Times: What should the Trump administration try to do with allies that was not possible over the last four years?
Mr. Estevez: Not possible? I think there’s lots that’s possible with allies, and I think we’ve done a pretty good job working with our allies around the –
Mr. Allen Yeah. So maybe revise the question to, you know, what is your advice here.
Mr. Estevez: Yeah. So my advice would be continue to do that work. It’s hard work. You know, allied interests don’t always align but from a national security perspective you pretty – find that there’s a good alignment, right?
When I go and visit allies, regardless of where they are, and start talking about the risks of artificial intelligence used by adversarial powers against, you know, not just us, them, they say, yeah, we need this to protect ourselves here.
And then, you know, they make their own trade balances, you know, and I don’t mean trade trade. I mean, like, where’s the line that, you know, they’re willing to press to and I think the – my advice to my successors in the Trump administration would be to continue that hard work. Foreign Direct Product Rule is a great tool in our toolbox but, you know, just willy-nilly using that is also not good balancing of interest there, right? Throwing Foreign Direct Product Rule on a country that is hosting U.S. forces that is probably at risk of adversarial missiles hitting their land may not be the best use of that tool, when you can negotiate an agreement with them.
Mr. Allen: So something that you said just sparked in me. You know, obviously right now one of the critical multilateral frameworks for export controls is the Wassenaar Arrangement. And the predecessor organization was COCOM. And one of the facts about COCOM, which was the Cold War era export controls multilateral arrangement – one of the facts that was for a long time classified but has since been declassified is that it actually was born as the economic adjunct of NATO.
And effectively, what the United States was saying is if you would like to be under the United States nuclear umbrella, if you would like to be under the United States security umbrella, you’re going to need to align with us on export controls. We don’t want you sending military relevant technology to the Soviet Union and then asking us to protect you from that same Soviet Union. And so I’m curious, you know, we talked about how Secretary Blinken has described this as the end of the post-Cold War era. And so I’m curious, you know, what do you think is the appropriate relationship between the United States, our security guarantees for allies, and our alignment on export controls with our allies.
Mr. Estevez: You know, this is – when we host a round table on this, and as a private citizen you want me to come back, I’m happy to, like, sit and talk about this for a long time. But here’s, you know, my quick take on that. You know, first of all, COCOM is a bad antecedent for this. There’s goodness in it, but economies changed from what they were in the Cold War, where I had this iron curtain over here, right, and the economies were not so intertwined, right? Obviously our economy is intertwined with China in so many different places, you know, including supply chains. And same applies to our European and our Asian allies.
The controls we put on Russia, frankly, impacted our European allies, who were willing to do it, way more than they did to us because they had a much more deeper trading relationship with Russia than we did. So it’s not about, like, you know, what our security guarantees are. I think, you know, our military alliances and our – you know, are critically important. I think they all recognize that. I also believe we need to sustain those alliances for our own good. Going it alone is not going to be an easy structure in any kind of real fight in the world. So that’s point one. But point two, so then you go on and say, OK, what do I need to control to protect ourselves? And I think they’re all willing to have that discussion with us.
Mr. Allen: Mmm hmm. That’s great. So this is sort of related. This comes from Ana Swanson of The New York Times.
Have you been in touch with the incoming Trump team? Do you have any concerns that a more unilateral, America first approach could damage the international coalitions you’ve been building against China and Russia?
Mr. Estevez: I personally have not talked to the incoming Trump team. There’s a team on the ground in Commerce who’s talking to the career guys, career officials, which is fine, about what we’ve been doing and how we’ve been doing it. The secretary certainly has talked to her potential successor, Howard Lutnick, on a number of occasions. I have talked to people who I’m pretty sure are going to be in key roles in the Trump administration, outside of, you know, official Commerce-dom. You know, I can’t say what they’re going to do. Many of them actually can’t really say exactly how it all plays out. (Laughs.) I think it’s – you know, my advice would be to sustain these alliances and build on them. And that’s the key towards true protection here.
Mr. Allen: That’s great. So if we can now go to folks who are in the audience, so my colleague, Brielle.
Q: Hi. Brett Fortnam, with Inside U.S. Trade. You mentioned that two more rules are coming out tomorrow. I was wondering if you could preview any of those for us. And I was also wondering, given, you know, the rule this morning, the rule yesterday, why is – basically, I’m curious as to the timing of these, why the rush right now? Is there a fear that the next administration wouldn’t pick up on the rulemakings, or that there’d be too much of a lag? And specific to the AI diffusion rule, I know one of the major criticisms is that there is a parallel processing that would allow China to basically get the same outcomes as it would be if it were able to get some of the restricted GPUs. I was wondering how that type of issue was taken into account.
Mr. Estevez: I’m not sure what you mean by the last one. I’m not aware of any parallel processing that would allow China access through any process that we have in that AI diffusion rule. Why rule? Let me answer the last part first, and then I’ll go to the first. Why pushing stuff out? Because, again, I’ve been in government in and out all my life. Transitions are hard times. We had a lot of stuff teed up. These are national security issues. Our assessment is that, you know, these are things that the new team – first of all, the new team, now, the AI diffusion one is 120-day period of discussion. May have plenty of time to make adjustments if they want to do it. It’s a massive rule.
But for most of these rules, there’s really a bipartisan view that these things are important. They’re national security issues. Leaving them hanging for a new team to figure out where the light switch is, how do I get in the building, where’s my PIV, you know, where’s my CAC card, who do I need to talk to about wanting to issue something, what’s the process? You can lose months to achieve the outcome. So, you know, just like I’m cleaning my desk out so that my successor will have a desk that they can feel is theirs and taking my own pictures down off the wall, I want to leave a clean slate of not hanging issues that they have to grapple with immediately so they can figure out where they want to go and do.
Mr. Allen: Yeah, I remember when I started at DOD – and this was the AI center of DOD, so, you know, not really like the – supposed to be the people who are slow in technology. I think it took me, like, three and a half weeks to get an email address. I mean, there’s just a lot of this stuff that stalls if you don’t keep your foot on the gas.
Q: Any preview on the two rules –
Mr. Estevez: Oh, the two rules. One will be related to biotech and the other is related to compliance around some of the –
Mr. Allen: Recent is-informed letters, yeah.
Mr. Estevez: I won’t talk about is-informed letters. But compliance around semiconductors.
Mr. Allen: I see. That actually, you know, you’ve jogged something in me which I wanted to ask. You know, if you look at some of the recent administrative settlements or fines that BIS has reached, there appear to be – at least based on the reporting in the news – you know, the fine is a tiny fraction of the actual sales that took place to China or elsewhere. And so I wondered if you could just sort of help us understand what is the right size for a fine, and under sort of what circumstances – like, how do you think about appropriately deterring these kinds of actions, while also rewarding companies who come forward willingly and divulge violations?
Mr. Estevez: Yeah. So let me go to the last one first. Obviously, if the company comes forward we give them all sorts of consideration on imposing, like, a breaking fine. Somebody gets significant fine, in cases – there was one recent one. I won’t name it, because I want to – you know, they self-confessed, and they worked with us. On fines for a company that we’re working through, first of all, depends on whether we thought we had a criminal case or not, which we’ve then gone through a criminal matter with the DOJ.
Mr. Allen: And this is – when you say criminal case, this is the knowledge and willful intent standards?
Mr. Estevez: Yeah, yeah.
Mr. Allen: Yeah.
Mr. Estevez: Second, you know, we do have some legal parameters under which we can fine, and you know what the caps are around that. And third is, like, is it worth going to court? So a $300 million settlement I thought was a pretty good settlement in that. You know, we started out at a bigger number. And that was, I thought, a pretty good number that we came out on, the Seagate fine.
Mr. Allen: Right. Great. Brielle, do you want to find somebody?
Q: Hi. Victoria Guida with Politico.
I was just wondering, how much do you think about the economic component of your work? And by that, I mean you framed everything in the context of national security, particularly as it relates to China. But we’re also, you know, in an economic competition with China. And AI and semiconductors are, you know, important sectors. And so I’m just wondering, is there also sort of an economic security component? How do you think about that in your work?
Mr. Estevez: Yeah, that should be an easy question to answer, but it’s not, because national security and economic security have, you know, a pretty good Venn diagram overlap points. My writ and the law that gives me authorities under ECRA is about national security, because my job is about focusing on national security and national security issues. So things I do are around national security, not trying to stifle the competition out there. Even when I do 232 tariffs, which fall under my writ, those are national security related tariffs.
You know, USTR just kicked off a 301 investigation on legacy chips. That’s a 301 investigation, not a national security, concern about dumping chips and, like, cutting – undercutting the marketplace on that. So that’s an economic issue.
The things we’re doing on cars are purely the things that I just talked about – the concerns of risks to your data; the concerns of turning your car either into a brick or, frankly, it could also be turned through software into a missile. You know, there’s a Julia Roberts movie where all these Teslas go run into each other.
Mr. Allen: (Laughs.)
Mr. Estevez: So, you know, that’s my job. That’s why I was asked to come do this job, because I have a national security background. And you know, my concern on the economic security side of that is, like, what’s the impact that I’m making.
Mr. Allen: Great.
Other folks in the audience who want to ask a question? She’s coming right to you.
Q: Hi. My name is Atsushi Sumikawa. I’m a master’s student at Georgetown University.
My question is about frontier model restriction announced just yesterday. As the U.S. private industrial AI producers are heavily reliant on foreign AI talents – H-1B holders from China and so on – to what – to what extent do you think enforcement will be possible?
Mr. Estevez: I think companies that, you know, want to remain in business are not out to violate the law and the regulation. I don’t see companies in their own self-interest wanting their model weights to be moved around the world unless you’re running an open-weight model such as Llama from Meta. So I think companies will do what’s necessary to protect their models.
Mr. Allen: Great.
We have an online question, and this will come as no surprise to you. This is from Karen Freifeld of Reuters: You came into BIS being grilled about export controls against Huawei not being strong enough. Since then, Huawei has only seemed to have gotten stronger. To what extent do you think U.S. policy and enforcement have failed?
Mr. Estevez: Well, I love you, Karen. (Laughter.)
Look, you know, controls are not about destroying companies, trying to put a company out of business. That’s not the way they work. Huawei is heavily subsidized by the state. State have used Huawei as their champion in a bunch of areas. So the state is going to support Huawei. So there’s nothing I can do to stop that from happening.
I think, you know, our work around slowing them down on 5G, which was the Trump administration policy around that, worked for a period of time, then it didn’t. The fact that they can put a seven-nanometer chip into a phone is not, like, a national security concern per se; it’s really, where is that chip coming from? They had, you know, a design house in HiSilicon who can design chips. I can’t impede that capability. I can’t impede where HiSilicon or Huawei was getting the chips in the Ascend 910B if they were getting them from outside of China.
Mr. Allen: And I’ve heard –
Mr. Estevez: And I think we’ve done a fine job in doing that.
Mr. Allen: Yes. I’ve heard that not just a majority, but a supermajority of all the Ascent 910B chips that have ever been made were made by TSMC, not made by SMIC, which I think highlights how the equipment controls have been effective at degrading SMIC.
Mr. Estevez: Yeah. And, you know, look, I’m not going to – TSMC, I’m known to them and has worked with us on stopping that.
Mr. Allen: Yeah.
Well, Undersecretary Alan Estevez, I want to thank you again for so much of your years of service both in BIS and in DOD, including those years that were given to you against your will – (laughter) – which was remarkable. But please join me in thanking Undersecretary Alan Estevez. (Applause.)
(END.)