Lawfare Daily: The Dangers of Privatized, Automated Immigration Enforcement

getting access to information that it used to rely on states for and it is getting access

“to technology that means that immigration enforcement requires less manpower to do the”

things that the human labor did before. It's the law fair podcast. I'm Tyler McBrine, managing editor of Law Fair, with Chini Sharma, an associate professor at Fordham Law School and contributing editor here at Law Fair. Under the auspices of national security authorities, we've seen kind of this system being

used to marginalize communities that have been marginalized under the banner of national security for a long time. Today we're talking about Chini's forthcoming law review article, immigration enforcement

intermediaries, and the automation and privatization of US federal immigration architecture.

So Chini, you have a forthcoming law review paper called immigration enforcement intermediaries. In it, you describe the current US immigration enforcement system in a few ways. I'll just name two of them, you call it an automated ecosystem of interconnected tech platforms. You also call it a code based Leviathan. So with those, those terms in mind, what is the current US immigration enforcement system

“that we have, and then briefly we can get into it more later, how do we get here?”

Absolutely. And like shout out to you guys, I think that the docu-series that you're hosting does a great job of like not just the tech infrastructure, but like broadly what is happening with

the relationship with the government and the private sector.

But I think there's kind of three pillars here, two that are like the traditional low guy of power in immigration enforcement, and that's a federal government in states. And then the new emerging note of power is private vendors. Private vendors have been involved for a long time, but their role has fundamentally changed. And it is not because we have generating AI all of a sudden, though that is how people seem

to think about it, but it is, and this would require, I think, more excavating of like the combination of socioeconomic, political developments that led to this, but vendors now

“have a very close, really tech vendors have a very close relationship with the federal”

government in a way that we haven't seen before, except for, perhaps the defense industrial and the military industrial complex. So what's happening here, we have a, and we call it an ecosystem, we call it a stack. It's essentially a re-architecting of how immigration enforcement works today. In that, once upon a time, it was a very manual process.

We have a federal government that has the onus of responsibility for immigration enforcement, and you have states, and states were necessary for effective immigration enforcement. Why? Because the federal government neither has the manpower to do what it wants to do or the information to do what it wants to do, states have robust information about their constituents.

They have more manpower, their closer to the ground. They can efficiently carry out certain activities that the federal government would rely on. And so the Constitution says that the federal government can't force states to help, states can't actively interfere with the federal government's integration initiatives, but they

don't have to help. And so there was kind of a tug of war of power, and yes, over time, more power shifted towards the federal government. And we can talk about kind of like why that shift happened. The state still had like a negotiating position, like a platform from which they could say,

we will cooperate in these ways and not in the use ways. And that has changed. And that is because the rise of these tech solutions that are now being invested in unprecedented amounts that we say things like exponentially increasing very casually, but it truly is increasing in a non-linear way, and because of the data and technology that these vendors

provide, the federal government is getting access to information that it used to rely on states for. And it is getting access to technology that means that immigration enforcement requires less manpower to do the things that the human labor did before. Of course, be incredibly ambitious and problematic deportation goals today will require quite

a bit of manpower, but a lot of what humans did before was replaced.

Companies that have contract with the federal government for a long time are continuing

to provide solutions, but now we have new smaller tech companies come in. And anybody who knows anything about government procurement knows one that it's incredibly

“boring, and two that it's incredibly important.”

But three, it's incredibly honest. It's very hard to meet the requirements to contract with the government, and it is also very hard to break in as a new entity because there are various things about how the government does business that it just wants to go back to trusted vendors. But the reason these new entities can break into the ecosystem is because there is this new

emphasis on interoperability.

Another on sexy term, but basically what that means is this apparatus is built to interconnect.

Everything that the private sector is providing is designed to play nicely with the other systems that other companies are providing. And what that does is create resilience in this automated architecture that we didn't have before, and it improves the capabilities that this architecture has because now a company is not just limited to what they have, both in technical capabilities and data.

They can benefit from the data sources and capabilities that other vendors are providing. So now we just kind of have the whole is greater than the sum of the parts, kind of a situation

of privatized automation and data streams that the federal government has at its disposal

that has supercharged immigration enforcement. Yeah, thank you so much for that actually very succinct answer to a monster of a question.

“I mean, this is one of the challenges I think that we were facing making deportation ink.”

The series is that it's just such a U.S. immigration system is so massive in terms of scale, number of actors, geographic scope, even that it can be just so hard to visualize a wrap one's head around. And so I think one way to get it that is we'll definitely dig into so much of what you just spoke about. But one way is to is to just take one vendor and maybe one technology and start there, which is where you start in the article with Palantir OS.

So could you tell me a bit about this particular software and its role in the immigration system? And then more importantly, why you chose to begin the article with Palantir? Palantir, I think a decade plus ago was kind of an, if you know, you know, cool new analytics company that was doing fancy Excel, not obviously fancy Excel, but just like very complex, mining of data to find insights. And that's kind of generally what people knew about it.

But as early as over a decade ago, or about a decade ago, Palantir was already working with DHS. These were contracts that were not getting the most attention. And that's because the backdrop of, you know, we talked about kind of how do we get here. One piece of how we got here is immigration got a makeover. And this is not to be facetious or make light of a serious thing, but post 9/11.

Immigration went from a, where kind of like, maintaining house, managing our borders, preventing contraband and enforcing kind of rough quotas that we have on entrance to the United States. And it wasn't something that got quite as much attention, I mean, not nearly as much attention as it's getting today. Post 2001 immigration became a national security problem. And really what immigration became because of the nature of 9/11 as the terrorist attack became an intelligence problem. It wasn't just that we have these overt threats to the country, but we actually have these very subtle sophisticated actors that embed within the country and have the capacity to cause very real harm.

Now, a different way to have looked at it is this is a national security issue that is related to a handful of individuals with certain ideologies completely unrelated to their status as immigrants. But what happened and there's a lot of great literature on why this is what happened is it became about the immigrants that entered the country. Very serious threat to us, hashtag not all immigrants, but enough that this is now something that we are not just monitoring it for quota purposes, we are monitoring it to make sure that nobody's a threat.

“And so when something becomes an intelligence problem, it becomes a data problem intelligence is data data needs to be in a format that is usable and you need to have the technology to be able to use it well.”

To volunteer enters the system, enters the ecosystem of immigration enforcement by providing something called the ICM that's the investigative case management system.

And what it is is basically a very sophisticated though now it seems rudimentary database that brought together a lot of data about immigration.

And so the purpose of taking a lot of data and putting it in one place and contracting with a company like Palantir that is known for and specialized in analytics is to take advantage of that data at scale to use the information to identify where the threats are and where you should prioritize limited government resources.

“So that's Palantir's kind of origin story here with immigration, immigration OS is like an enormous level up.”

We have now evolved past ICM to something that can be characterized as like a new layer of technology on top of ICM.

But what this new layer of technology does is it kind of opens up new pipelines to new data sources that did not exist before. For data sharing from other parts of the government, including states and localities, but we also have a technology layer on top of ICM that Palantir has built that says we can actually take in data from all of these tech vendors as long as they use the data standards or data formats that we require.

“And once we can do that, we are just funneling in new information that we can use our advanced analytics software on to identify the insights that the immigration enforcement regime wants to carry out at schools.”

This last thing I'll say is like what's the role that it's playing? We in our paper call it a hub and spokes model and Palantir and immigration OS. But really just like the suite of centralizing technologies and interoperable technologies that Palantir provides is the hub. It can be visualized as like the part of the wheel that combines all of the spokes and the spokes are these other data sources. Many of which are private technology vendors. And so the reason we start with Palantir and the reason I imagine will end the conversation with Palantir is because it is the one piece of the system that feels fundamentally integral.

The DHS procurement documents around Palantir are actually confirmed that they have taken various steps like single source contracts and what a single source contract is is basically we're not entertaining other bitters. We are going with this vendor, but we have to justify it because generally contracts need to be competitive and we're justifying it by saying no other company can provide us this set of capabilities that we need at the cost and at the speed that Palantir can.

“So it is this like keystone to the entire enterprise.”

As you mentioned Palantir used to be and if you know you know kind of a company and now of course it is in the headlines it is something of a specter to certain parts of the country people concerned with surveillance and civil liberties, constitutional protections, etc. But one insight from your paper that I thought was really interesting was that sort of make this argument that yes these these very powerful technologies can be very scary in this regard and cause all these harms. But a lot of times they oversell what they can actually do and there's a true dystopian worry here that it's actually in these errors and mistakes that even more harm can come and and this is not just hypothetical it's it's happening.

Could you talk a bit about that because that really kind of you know jumped off the page of me on one hand it is true that we have technologies entering the space that are doing things that the government could do before.

The other hand and this should be a surprise to nobody and not considered at all a controversial statement.

The private sector over sells that's part of a business model especially in the tech world where often like your minimum viable product just becomes your product because it is the thing that somebody will pay for right and especially in the government contracting context. The government once it invests in a technology system becomes independent on it because it builds the rest of its institutions around that system and so then they become legacy systems. That just kind of get auto renewed over time because it would be very honest cost prohibitive even to swap it out for another alternative and the reason all this is relevant is it takes away more incentive from the private sector to continue to deliver the highest quality most cutting edge products if you know that you kind of have somewhat in elastic demand from your main customer.

Why is this a dystopian reality? It is an offer world in which we have so much data that we can perfectly accurately identify individuals who might be violating immigration laws identify exactly where they are and bring automated enforcement actions against them. That is scary for many reasons that we can talk about but it is even scarier to have systems that are shortly built prone to inaccuracies and as we talked about when there isn't as much incentive to go above and beyond in the responsible and careful design assistance.

“We know very little about just how inaccurate these systems are and that's because very very real violence is being enabled by these systems.”

For example, volunteers built in the system called elite essentially what it does is it aggregates information including tips from hot lines to ice and taking all of that information it identifies hot spots of areas that are going to have a large number of individuals that are of interest to federal immigration enforcement. And for every individual will pull up dossiers about that individual including photo address phone number place of work other people living in the household even social media content and some of this data is like provided by companies like Thomas Reuters.

Why is that scary that information could be wrong and now you have masked ice agents sometimes not clearly presenting as ice agents knocking on doors infringing on individuals liberty private spaces potentially using this information pretty textually to go after individuals that they don't actually believe to be violating immigration laws. with faulty information. So you have individuals that are being detained wrongfully you also have kind of like products that ice agents are trained and directed to treat as authoritative and why is it concerning when that's faulty for example that are there are apps out there that are meant to identify whether or not an individual is legally within the country and what their status is as an immigrant.

And so you can have individuals there are cases that are reported about this individuals presenting birth certificates like paper documents to the ice agent and the ice agents like my mobile fortify app has said that you are in fact lying to me right now and I have to trust the app and like what that.

“The combination of inaccurate systems with what you see in that example is there is long documented behavior of just deferring to outputs from systems we think or believe they're more accurate.”

We tend to want to delegate judgment. We think these systems are neutral especially when we don't know the error rates but beyond that if you think and there's reason to believe there's nefarious intent in these actions. They provide great cover for something that you would otherwise want justification to do.

Hey there listeners I'm Jen Patia director of audience engagement for Lafaire and I mostly work behind the scenes to edit the podcast you're listening to right now.

Today I'm stepping out from behind the editing desk with a special message. I hear from listeners all the time about how much they rely on our daily episodes and if you're one of those listeners I'm here to ask you to help us keep it going. Every day I get to see up close what it actually takes to make this show happen. There's a lot of care and effort that goes into producing reliable and dependent nonpartisan analysis on a daily basis and right now that work feels especially important.

“Things are complicated the stakes are real and cutting through the noise isn't easy but that's what Lafaire tries to do makes sense of big national security and legal questions without”

Deming them down or spinning them up and we keep it all free because we believe that access to quality analysis and independent media shouldn't have a paywall.

But to keep doing this work we need your help. Lafaire is a 501c3 non profit and we rely on the support of our community of listeners and readers to make our work possible. So if you tune in regularly and this show has helped you feel a little more informed or a little less lost in the chaos consider becoming a monthly supporter.

Thank you for listening for caring about the things that matter and for supporting Lafaire.

“Yeah, I mean the words or wellian and Kafka ask are thrown around too much nowadays but these are situations in which they seem fairly apt.”

But I want to take a step back for a second and throw up a counter argument which you've already started poking holes in anyway but which I want to hear you we could take it in a few different directions but and you address this in the paper but. The counter argument could be the government routinely outsources core capacities to private vendors sometimes to to great benefit to to citizens and the people government serve and in in terms of immigration enforcement especially the you know the American people voted for this administration who campaigned on these immigration numbers that the government under its current capacities could never.

And so they have to contract out to private vendors what is wrong with everything I just said or right about what I just said.

I mean we could we could take it in the direction of why it's you need the unique harms and the immigration context or the lack of democratic oversight and the capacity of the system. But you know you you also address these these kind of arguments so I just wanted to open the the forest of the app. No, I'm so glad you did and I am going to try to be coherent and organized and responding that it's one of those areas where there's just so many different ways to approach that line of logic. And it kind of depends on your priors so I don't think that I've really hidden the ball on my views of the immigration enforcement system. I think that you can say. I believe that we should have stringent immigration enforcement.

“That's why it's being carried up today is unacceptable under various legal principles both in technical legality but also in like fundamental principles about how we want the lot operate in the world.”

Now, if that's the case then what we are seeing is the delegation of a critical executive function to private vendors.

A relationship is not something that's happening through bills passed by our representatives in Congress. It's not happening through agency action that subject to public notice and comment. It is happening through contracts. It is happening through contracts that I'm not sure that there are people out there that care about the D far or the far, which are the enormous statutes that regulate how you can contract. And so what this means is we are shielding these policy decisions from public accountability.

There is not kind of public awareness of how this is happening. There is not public input into the terms of the contracts. And then you can go back to, well, yeah, we picked an executive. We picked an executive. Courts have said that actually the executive has enormous power should not be incumbent in its use of that power.

“And it has a lot of discretion to discern how it wants to take care of the faithful execution of the laws, right?”

All of that might be true, but you might still believe. And this is a separate paper that I'm writing because I'm clearly existential like preoccupied with the executive branch is use of automation. That even if we have a unitary executive with this enormous amount of power, we still believe that there are other things in the government that provides some friction in execution of a president's will. In a way that maintains the legitimacy of that use of power. And some of those things are informing the public, involving some degree of public participation, incorporating some amount of like human judgment in the way something is executed.

What we have when we use systems instead of agencies to carry out what an executive wants is there is no longer any risk that you're going to have individuals not comply or exercise discretion in ways that are contrary to what the president wants. Or even have the knowledge that they would get from being the on the ground enforcers of a law or a policy that would lead them to potentially whistle blow cooperate with Congress share something with the media even leak information, resign and protest because so much of this is happening as technical specifications in esoteric contracts that are made between the government and private vendors.

You now have just like these many layers of things that are technically laws and doctrine that further shield what's happening from public accountability.

And so the justification for a unitary executive is democracy put that person in power and so democracy should have something to say about the way the power is used. And so if the justification is we put that person in power, then we should have the information and the ability to provide input to eventually decide we no longer want them to be in power. And I'm explaining is the way this is working is denying the public that information. And then because this is a valid viewpoint that I don't think people should be quite so hesitant to stand behind is you can think we are wrong about the way we are approaching immigration.

“But it is not the national security issue that we have painted in broad strokes to be that that is not the humane or autonomy respecting way to have go about maintaining our borders.”

We can agree that there might be some amount of control that has to exist at borders but to paint the broad stroke of every potential immigrant in the country, whether they want to enter the country or already in the country is going to be seen as a potential national security threat has led to a way of thinking about immigration.

That makes the problem even worse because if you have the capacity to go after 10 people and you have 100,000 that qualify for individuals that you can enforce against.

You have quite a bit of discretion over the 10 people you go after and there is no shortage of evidence that those 10 people aren't picked randomly and so under the auspices of national security authorities we've seen kind of this system being used to marginalized communities that have been marginalized under the banner of national security for a long time.

“One more question on the harms because I think we've already painted the picture of quite an intractable problem but I think one of the most worrying aspects is this point that you make that.”

These surveillance technologies that are being developed in the immigration context are not limited as such and in fact these companies are incentivized to grow their business and expand it beyond the immigration context and there's no shortage of capacity problems elsewhere in government. So this could be generalized could just be a bit about that danger whether speculative or already here in happening. This is a concerning one that is unique to not just automation but also the role of private vendors as providing the tools and data that the federal government is using and so I'll kind of go through two ways that there's scope creep here.

In the federal government procure systems they kind of get the first cut at this is what we want this is how we want it to be built these specifications we have.

Specifications are not kind of neutral directives on system design they are baked in policy determinations like.

“What is a threshold that triggers some notice or kind of action one great example and by great I mean horrific example is I believe it was about 2017 though the reporting happened much later.”

It came out that at the executives behest a automated tool that recommends whether or not an individual should be detained when found to be. Was changed overnight without any public awareness again this reporting came out years later to categorically deny recommendations that somebody not be detained. So under law you have ice agents that are provided with the discretion to say this individual doesn't need to be detained but when they turn to rely on a tool. The tool is designed in a way that that discretion is no longer operates on the ground.

Now how does this lead to scope creep well there is a common problem well reported that states and localities don't have the resources to do what they need to do.

And when you don't have resources modernization or automating things becomes incredibly tempting.

“Now states are not going to go out and say Palantir build me a new bespoke system that serves my approach to policy determinations around how we're going to cooperate with the federal government.”

What is going to happen and what is happening and we've seen this in the incremental law enforcement context through products like Gotham is that they're going to buy off the shelf products and what are these off the shelf products. They are packaged versions of what was built for the federal government and what does that mean when states and localities by these products they are potentially completely unknowingly embedding into their systems the policy determinations that were made upstream in designing the products for the federal government.

So that's one kind of scope creep states are now kind of roped into this mission even if they're doing it unknowingly.

Second scope creep we should all care about this issue because we should care about immigrants.

We should also care about this issue because immigration and the way we treat immigrants in this country has often been a canary in the mind field of what is to come for the rest of the population. It's something that and this is something that automation does and we've seen with both federal government surveillance and commercial surveillance. It normalizes and desensitizes the public to certain kinds of activities we are no longer. We're done by the idea that ALPR cameras are tracking our license plates wherever we go collecting this data and able to monitor our movements because it's been around for a while now ALPR data.

Many contracts say we are only sharing this with federal immigration enforcement under these terms and conditions however those are limitations put in contract those contracts can be changed.

They don't even have to be changed there's also many documented cases of even for example sanctuary cities that say we are actually categorically not sharing this data with federal immigration enforcement because we do not want to cooperate in this regime. What is that dependent on? It depends on individuals complying with that and there are many cases of individuals sharing that information with ICE agents anyway. So all of this is the restrictions on a technology once it is built and out there feel very thin and the incentive for both the private sector and the government is to expand how they're used.

The federal government doesn't just want data for immigration enforcement they want data across the board for law enforcement for intelligence and national security even when it doesn't relate to immigrants.

“That is the interest that's why we have the fourth amendment because we think that that's going to be the default position of the government.”

The private sector predisposed or inclined to expand where they use this that is more business those are more contracts that is more profit and so now we have the private sector and the federal governments incentives aligning and that's a scary thing for the public. And so I think it is very likely and we are kind of already seeing this that these technologies start being used outside of the immigration context. So again you have painted I would say a pretty bleak picture probably both for those who favor fidelity to our immigration enforcement laws and those who are who will support disagree with them entirely.

I mean there's so many of what we just talked about the entrenchment of the technologies and the companies now profit can drive policy decisions how this sector is insulated from oversight. The other branches there's reference now that it's been secretized but your paper also gets into recommendations for interventions or proposals that perform and and not of the sort that are you know typical of what you would hear in like a very unsexy procurement. The context the audience largely is if I'm not mistaken the states and and what can be done within our system of federalism specifically so turning to the end here.

What do you propose given the just the scope of this. I hesitate to even call it a policy problem because it's a systemic it's it's a system wide. Yeah so if you could lay out some of those interventions.

“It's important to paint a realistic picture something so you know what you're up against and it's also important to recognize like that's not the end of the story.”

What do we do if you just give up in the face of that and states have actually been kind of like. The bastions of hope in this context in that they are doing what they can and sorry not just states sub federal entities you have localities resisting their state and the federal government with a state wants to cooperate with the federal government.

These sub federal entities are resisting have resisted in the past I mean.

“Even before this enormous automation apparatus we had sanctuary cities and I think that they are.”

I mean immigration enforcement started with the story of federalism. The federal government relied on states to some degree in states had a negotiating position with the federal government on how much they want to help and how much they don't want to.

To help and I think it comes back to federalism of states do have their own spheres of power and within those spheres of power they are capable of protecting their constituents.

They can proactively ensure that there are protections for their constituents that go above and beyond what the federal government and how the Constitution is being interpreted would give those constituents. So it can be everything from banning the use of things like biometric surveillance and facial recognition by law enforcement. And the reason we talked about kind of the scope creep once law enforcement has it they can share that information with DHS.

It doesn't have to be something that was only collected in the immigration context.

“And so banning facial recognition and biometric surveillance is at least one way in which you take away like a big key that the federal government uses to match individuals to other sorts of identifying information to build these dossiers.”

Now this is a step. It's an important step. It's a small step. There is so much other information that's being collected like you have information from utilities bills you have information from your grocery loyalty card and what you tend to buy and the amount of insight you can get from that seemingly innocuous information is enormous. So you can also have enshrined in your state constitutions increased protections for privacy above and beyond what the Fourth Amendment might provide. You can have privacy protections that you say within this state we actually extended to immigrants in the way that the federal constitution has not been interpreted to extend to immigrants.

They can be explicit about the kind of digital privacy that is protected. There are also conversations around Ken states through constitutions or statutes say that we are going to provide immigrants the capacity to bring bivins actions and why is this important just kind of not going into the legality of it.

“Once upon a time doctrine had been interpreted in a way that makes it very hard for immigrants to bring a lawsuit against the federal government saying you have violated my constitutional rights in your actions against me.”

And now states are using theirs fears of authority to say we are granting the constituents of this state including immigrants the capacity to sue for violations of rights. There's a lot of nuance there. There's a lot of conversation in the scope of that but that is something that is being pushed. I mean I can go on about the different structures that can be set up but like what states can do is to go beyond contractual restrictions on just sharing of information. Because as we talked about those are paper thin they rely on human compliance and they are also leaky so for example a state might have a contract that says do not share this information with ice but the state also sells this data to data brokers who then solve the data ties or you can have states that make this information available on their websites so that they can then be scraped and then made available to ice.

There's just a lot of leaks in that and so what states can do and what some are doing is dismantling the apparatus itself. Take out the pieces, the spokes to the hub model, these data streams, the use of these vendors that are known to cooperate with federal immigration enforcement, cut off their access to information about your constituents do not provide the business services that they might rely on do not use these off the shelf products. It's not just law passing but about proactively dismantling the architecture that's allowing for this.

One thing that I just love to give a shout out to an entity that was not called out in the paper but is a form of resistance that we had not thought of but is really heartening. There is a Canadian Union BCG EU that is a minority shareholder in Thomas Reuters and since 2020, but this has been reinvigorated in a huge way recently, they are using their rights as shareholders to try to unearth information that many suspect that Thomas Reuters is withholding are about its cooperations with ice.

We do not cooperate with ice. There is reason and reporting to suggest that is not the case.

“And so now these shareholders are saying we have the right to make proposals and the proposals that they're making are for human rights impact assessments into Thomas Reuters business activities and books.”

And what that will do is it will uncover information that has otherwise maybe through litigation or investigative reporting previously been unknown to the public.

And so even in the corporate contacts, there are things to do to resist to signal discomfort to say I want to change company practice from the inside as a shareholder. As you've said, there are so many different directions we can go even from here, even after having spoken about it for 45 minutes, but a reminder that this is a deeply entrenched but not entirely insurmountable problem that seems like a good place to close.

“The paper was really important, really thought provoking. I hope it's the first of a series of papers exploring this research agenda. So, Cheney, thank you so much for joining me.”

Thank you so much for having me. The Lawfare podcast is produced by the Lawfare Institute.

“If you want to support the show and listen ad-free, you can become a lawfare material supporter at lawfaremedia.org/support.”

Supporters also get access to special events and other bonus content we don't share anywhere else. If you enjoy the podcast, please rate and review us wherever you listen and really does help. And be sure to check out our other shows, including rational security, allies, the aftermath, and escalation. Our latest Lawfare presents podcast series about the war in Ukraine. You can also find all of our written work at lawfaremedia.org. The podcast is edited by Jen Pazia, with audio engineering by Kara Shillin of Go Brodyo.

Our theme music is from Alibi Music.

As always, thanks for listening.