DOJ vs Data Espionage
Data weaponization in the 21st century
The DOJ is now charged with protecting American data from foreign adversaries. This new proposed rule they recently issued is, according to one observer, “one of the most ambitious and sweeping new initiatives in national security law over the past few years.”
To discuss, we interviewed Devin DeBacker and Lee Licata of the Department of Justice’s National Security Division.
We get into:
How adversaries plan to weaponize obscure data types — including geolocation data, DNA sequencing, and undersea cable transmissions;
How China managed to purchase genomic data on millions of Americans through healthcare investments;
Why black box data brokers keep records of who goes to casinos;
How the DOJ plans to protect your data, and whether their plans can be thwarted by gridlock in Congress.
I’m excited to introduce a partnership with Policyware to bring affordable, expert-driven policy education to my audience. Starting May 14, Samm Sacks will be teaching a deep dive into China’s Digital Governance and its Global Implications.
Samm is an old friend of mine and a Senior Fellow at Yale Law School’s Paul Tsai China Center. She is a leading expert on China’s cybersecurity legal system, the U.S.-China technology relationship, and the geopolitics of data privacy and cross-border data flows. Check out below a show I did with Samm on ChinaTalk discussing China’s digital governance.
You’ll learn over several weeks as Samm delivers live classes, with options to listen on your own time. Policyware Deep Dives are designed to be attended alongside your job, and they will help you organize with your employer for cost sharing. Check out the show we did together on data issues late last year.
Help support ChinaTalk by registering for the deep dive here and thank you to Policyware for sponsoring today’s episode.
The risk of data weaponization
Jordan Schneider: What does this proposed rule do?
Lee Licata: The executive order allows the DOJ to set up a new program that will prohibit or restrict certain types of commercial transactions between a US person and a country of concern or a covered person — which is someone who is owned by, controlled by, or subject to the jurisdiction of a country of concern — where that commercial transaction would afford the country of concern access to either bulk sensitive personal data or government-related data. Now, the regulation itself sets out two prohibitions and three restrictions.
The two prohibitions are on data brokerage. In other words, the sale or leasing of access to data on a US person to either a country of concern or a covered person, or transfers of genomic data, which we see is one of the most sensitive categories of data. Both of those will be prohibited in the initial regulations.
The other piece of this is the restrictions. Restricted transactions could be allowed if the transaction can be commenced with certain security measures in place, such as data anonymization.
There are three restrictions — investment agreements, employment agreements, and vendor agreements. Those are transactions that do have some national security risk, but that risk may be lower relative to the economic value of those commercial relationships.
Devin DeBacker: Bottom line: what we’re talking about are counterintelligence risks — the ability of foreign adversaries to take some data and use it to influence someone in the US to do something they want. Data can provide leverage to a foreign adversary, or allow them to identify useful people to begin with.
Jordan Schneider: The list of countries is not super surprising — Venezuela, Iran, North Korea, China, Russia, Cuba. But the list of specific data types is quite interesting — genomic data, biometric data, personal health data, geolocation, financial data, and certain kinds of personally identifiable information. How did you decide what to include and what not to include?
I want to emphasize that there’s a huge difference between a national security regime like this and a privacy regime. Privacy regimes like the one in California cover a broader, longer list of data types.
Regulating expressive data is not the goal. Things like text messages, emails, academic publications, and music are not covered under this program.
Lee Licata: The rule wouldn’t cover the phone book, as we like to say.
Jordan Schneider: I enjoyed the fact that you put Lee’s phone number into the proposed rule so that people can call you if they have any complaints.
Lee Licata: That’s right! That’s part of a transparent rulemaking process. We do need to make ourselves available to hear input from the public.
Jordan Schneider: What’s so scary about the specific types of data included in the new rule?
Devin DeBacker: I’ll talk about genomic data first because it has some unique characteristics. We think this category is the most sensitive out of the data types because it’s the one with the lowest proposed bulk thresholds.
Genomic data exposes vulnerabilities in people and populations. Genomic technology that is used to design disease therapy can also be used to identify genetic variability in a population. Variations can be linked to susceptibility to disease, which can be used for good purposes as well as nefarious purposes, including bioweapons that target those vulnerabilities.
There are large genetic data sets that are used for tracing ancestry, solving crimes, and research. Those can be misused for surveillance and for repressing certain segments of society.
Matt Olsen, the assistant attorney general, calls this threat “Transnational repression.” Those genetic vulnerabilities in an individual or a population can be used for the same sorts of targeting, counterintelligence, and blackmail intimidation.
The reason genetic data is so particularly sensitive, though, is that it’s persistent in a way that your bank account PIN and your IP address aren’t. Those can be easily changed, but you can’t change your DNA sequencing very easily.
Jordan Schneider: Wow, uh, that’s scary.
I made a joke during the Biotech 101 show that someday terrorists will be able to create a designer drug to kill Jordan Schneider and no one else.
But evidently, you have information that leads you to believe that targeting big swathes of people based on their race is a real concern.
Devin DeBacker: I’ll say one more thing — the IC has been warning about this for a while. They put out a series of bulletins, including one that detailed China’s track record of collecting, analyzing, and using genetic data to track and repress the Uyghur community in China.
It’s pretty worrisome.
Lee Licata: We’ve already seen Chinese companies making investments in the US market, particularly in healthcare, in companies that have access to this kind of data.
In 2013, the Beijing Genomics Institute (BGI) bought the US company Complete Genomics, which had DNA sequencing on millions of Americans. In 2015, WuXi Pharma bought the US firm NextCODE in similar circumstances.
We struggle to think of a legitimate commercial reason for a foreign-adversary-based company to have access to millions of Americans’ genomic data.
The Chinese have exploited our open investment climate and made investments in the health sector that give them access to this kind of data.
Jordan Schneider: I can imagine that this might cause some scary situations during a conflict, in which an adversary could use to AI generate calls that appear to be from the mothers of key personnel, saying that there’s a family emergency and that they need to desert immediately. [Editor’s note: Russia has already deployed such tactics in Ukraine.]
Lee Licata: Correct. The ability to use big data analytics creates a far more effective and targeted means of conducting espionage and blackmail.
The ability to use technology to find the needle in the haystack has just been transformed over the last few years. But a lot of the data we’re finding can provide a really clear picture of people’s patterns in life.
For example, Duke did this study where they tried to see what sort of data on military members, veterans, and government employees they could buy from data brokers. They found about 8000 data sets related to those types of people. Those people may be valuable assets because of where they work or the information they have. But some of the things they could find in those data sets were really revealing. You could buy access to data about how often a person frequents casinos. Combine that with some financial data about someone and what debt they might have, that’s a really interesting data point for a foreign intelligence service.
Jordan Schneider: What about location data? Why do you guys care about other countries knowing where people are?
Devin DeBacker: There’s concern about foreign adversaries targeting, identifying, and tracking dissidents.
But when we talk about geolocation data, it goes beyond just tracking dissidents.
We are concerned about aggregated insights. When the Strava heat map was published, researchers were very quickly able to take that information — which in theory was only relevant to fitness and exercise — and use it to map government facilities and military bases.
That’s just one data set from one company. You can imagine the kind of aggregated insights that would allow a foreign adversary to have if they had many data sets or a broader slice of data. It’s both a micro-targeting concern and a macro-targeting concern.
Subsea cable coercion
Jordan Schneider: What’s the 101 on Team Telecom?
Lee Licata: Team Telecom is an interagency body composed of representatives from the DOJ, the Defense Department, and Homeland Security.
That body is tasked with reviewing certain types of FCC telecom licenses that have requisite foreign ownership or a foreign nexus for national security and law enforcement risk.
Over the last few years, though, probably the big focus area of this committee has been submarine cables, and it’s articulated in this EO. This is a space where we see a lot of this kind of data risk as we look at the evolution of the cable market.
Jordan Schneider: What is so interesting about undersea cables?
Lee Licata: 95% of the world’s Internet traffic now moves through a subsea cable. They’re an extremely important piece of infrastructure for connecting the world digitally. But here’s what’s happened over the last few years.
Traditionally, cables were owned by telecommunications companies — the Verizons and AT&Ts of the world — as a means of moving their traffic between networks and different continents. But over the last few years, what we’ve seen is that cables are now part of the cloud.
What we’ve seen is the cables being built by the large hyperscalers and the cloud operators to connect their data centers on a global scale. In other words, to move lots of data around their architecture. Now, most of the cables that Team Telecom looks at are being built by Google, Meta, Amazon, or Microsoft.
There’s a very different national security risk profile that comes with those kinds of cables, both in terms of the sensitivity of data that’s running over those cables and the volume of data that’s running over those cables. This is a piece of infrastructure that we weren’t really thinking about a decade ago, but now we see it as a critical node and a vector for this kind of national security risk.
Jordan Schneider: Why is a Meta cable more worrisome than a Verizon cable?
Lee Licata: Because of what moves over it. When those companies are building cables to connect data centers, what they’re really doing is moving their back-end data around the world. All your text messages, Facebook data, Instagram algorithm information, WhatsApp communications, and the documents you keep in the Google Cloud — that’s what those cables are moving.
It’s a different type of traffic that has more sensitivity in our view, and the volume is just wildly different. The cables that we’re seeing built now are starting to bend the laws of physics in terms of the amount of data that can run through them.
Jordan Schneider: It’s not just Western firms building cables, Huawei has also announced lots of China-to-Africa cables. How can these cables pose a threat?
Lee Licata: We usually think of two things — landing location and ownership. Now, a lot of cables are owned through consortia. It’s not one company that owns the cable, but it’s a group of companies that all make investments into the infrastructure, and then they each own a part of it to move their traffic.
For example, the PLCN (the Pacific Light Cable Network) was originally proposed to land in Hong Kong as well as Taiwan and the Philippines. One of the owners was a Chinese telecom company called Dr. Peng, alongside Google and Meta.
For us, when we look at that risk scenario, landing in Hong Kong was very high risk because the Chinese were re-exerting their influence over Hong Kong at that time.
The second concern was having the Chinese company in the ownership consortium, which would allow them access to all of the cable’s infrastructure. Ultimately, we recommended to the FCC that the cable not land in Hong Kong, and to get rid of that one owner, which the consortium did. Then they came back to us with a new cable application that only comprised Google and Meta, and we then entered into mitigation agreements with both parties.
Devin DeBacker: A couple of years ago, the ARCOS-1 cable was proposed to go from Miami to Cuba. That was the first outright denial that Team Telecom recommended.
Lee Licata: As we see in many of these cables, the owners of the cables can lease capacity and allow others to use them. What we articulated in the recommendation was that the ultimate customer for the cable to Cuba was actually the Cuban government, not the applicants for the landing.
The executive rule-making process and public input
Jordan Schneider: In the ANPRM phase of regulation, people get to read the lovely 30-page document you published and send you comments, which is a neat process. This document also included a list of questions that you have for the public.
Would you like to highlight some open-ended questions to our readers at home?
Devin DeBacker: One of the big questions is where to set the bulk thresholds.
The program applies to sensitive personal data related to government personnel and locations, and the threshold for that is zero.
But we also care about bulk data on non-government personnel. That’s because, in sufficient quantities, those datasets could be used to draw aggregated insights, as we mentioned before. The question is, what is that quantity? For geolocation data, would info on 10,000 US persons per year be weaponizable if it were in the hands of a Chinese or Russian company? For personal health data, what number of people would it take to draw dangerous insights? One million? What is the right number?
That really doesn’t have a lot of precedent.
It has to account for risk, as well as economic impact. It has to account for smurfing concerns, which is a technique for circumvention by lumping together smaller transactions over time to get the amount of data you need. There are other types of circumvention as well. That’s one area where we’re very interested in feedback.
Lee Licata: Another area I’ll highlight is the impact on the private sector.
In the ANPRM, we outline a menu of both physical and logical access controls, as well as privacy-enhancing technologies and data anonymization techniques. So the question is, how effective can those controls be at protecting a data set in one of these transactions? How feasible is it for companies to deploy such measures? How much will it cost? We would like input from the private sector in this area.
Jordan Schneider: Why was the Department of Justice tasked with this instead of the FTC or some other government organ?
Devin DeBacker: This program is a national security authority addressing counterintelligence risks that live in the private sector. That’s what the DOJ does. We do it through CFIUS, we do it in Team Telecom, we do it in civil and criminal enforcement of export controls and sanctions.
The DOJ’s focus on corporate compliance and our expertise in data security made us the right choice for the job.
We’re going to have to create a new unit here in the foreign investment review section, which Lee is the inaugural head of. We’re going to have to hire a significant number of attorneys and non-attorneys. We’re going to have to have a team responsible for licensing. We’re going to need a team responsible for targeting and designating covered persons and covered entities. We’re going to need a team responsible for regulatory development.
We’re going to need to expand our team of subject matter experts, and we’ll need a compliance and enforcement team.
But we’re going to need a significant expansion, and that expansion has already begun.
Jordan Schneider: Do you have to sell yourselves to Congress? Who decides if Lee gets staff?
Devin DeBacker: Yes, this will need new money. We’re working with OMB and Congress to make sure that we can secure the resources necessary to bring this program online.
Jordan Schneider: What happens if you don’t? Do we just not get the rules?
Devin DeBacker: We’re going to continue in the rulemaking process.
The next step is issuing the NPRM, which we’ll do later this summer, and then we’ll move forward on the rest of the program, including once that comment period is over, we’ll have a final rule out. When the comment period for the NPRM is over, we’ll begin thinking about what our initial lists of designated cover entities look like and so on and so forth.
Let’s not put the cart before the horse.
This is an area where there’s a lot of bipartisan and nonpartisan support. There’s been a recognition across administrations, across parties, and across houses of Congress that the data threat posed by foreign adversaries is real and we need to secure ourselves.
Bureaucratic make-believe and regulatory reality
Jordan Schneider: If you two had a magic wand, what type of authority would you grant the DOJ to implement a program like this?
Devin DeBacker: IEEPA is a pretty good authority. Now, it’s not a dream authority because it’s an actual authority, but making an IEEPA for the 21st century would be nice.
Actually, the dream would be an update to any of the authorities we already have. Take the Cable Landing License Act for example. That was passed in 1921. It’s a hundred years old. We’re regulating hyperscale submarine cables with fiber optic connections using the telegraph authority.
Lee Licata: We’re simple people. My wish would be for congress to codify Team Telecom in a statute instead of an executive order. Then I think there’s definitely something I’d like to see the ICTS authority built out, which may reflect a few years of trying to get that authority to work at The Department of Commerce.
Devin DeBacker: There’s certainly room for Congress to compliment what we’re doing. I mean, what we’re doing in a lot of different avenues in the government.
With attempts to regulate things like Wechat, for example, we hit the limit of some of our existing authorities. The cable law was passed in 1921, IEEPA was passed in 1977. There’s a lot of room for Congress to update these authorities. That’s my dream.
Jordan Schneider: What a sad answer. Oh, man.
It seems like a pretty tight talent pool. What is the process to attract qualified people and upskill the ones you have?
Lee Licata: Technology experts are certainly one of the things we’re looking for. You’re right that there’s a limited number of lawyers with tech expertise.
But we’re also looking for more rule writers. We’re looking for people who have expertise in the Administrative Procedures Act. We need folks who have expertise in things like export controls and sanctions, who know how designation lists work and how we litigate those designations.
We’re not looking for one type of lawyer here. There’s a whole array of talent we’ll need.
Devin DeBacker: Our strategy has been to bring in people with a very diverse set of backgrounds and experiences. It’s been very intentional and it’s worked very well. We’ve got folks who are former State Department employees. We’ve got DHS, we’ve got IC folks here, lawyers, we’ve got White House folks. We’ve got litigators, we’ve got prosecutors. That’s, I think the key here, right, is because all the problems that we’re trying to solve are diverse and they’re complicated and they’re complex, and they have a lot of different perspectives that you have to bring to bear in order to solve them.
We bring a lot of different perspectives into the office. Maybe one of us isn’t enough to solve this problem, but all of us together can solve it.
Jordan Schneider: This was unfortunately not a sponsored episode by the Department of Justice’s HR department. But I guess if you’re interested, look up the proposed rule, find Lee’s number, and go have a chat. You might get hired, as long as Devin can sweet-talk the right congresspeople!