Transcription
Probably the most controversial thing that Palantir software is being used for right now, I would say, is for immigration enforcement. Um, with ICE and other, other organizations. Can you talk a little bit about how Palantir software is integrated there?
Yeah, certainly. Um, well, first, I think a lot of the controversy moves with the politics. You know, we first started working with ICE when Jaime Zapata, an ICE agent, was assassinated in Mexico in 2011, I believe it's either 2010 or 2011, um, under the Obama administration. And we've worked with ICE from then on till the present moment.
Um, the, the work is, is structurally very similar. It's like there's information that DHS has access to that are spread out across the component agencies that, um, humans are trying to look up. They're literally doing searches in all these systems. So how do we enable them to actually see the whole picture and then use that for resource allocation? So where do we have threats? Where should we be allocating people? How do we prosecute, uh, removal operations? Um, that's, you know, in many ways, exactly what we're doing.
So a couple things, right? So number one, I think Palantir, you have a $30 million contract, um, to build a platform called Immigration OS. So what kind of data that's available to enforcement and removal operations?
So encounters at the border, right? Um, asylum applications or the lack of one, um, applications for benefits. Uh, so a lot of this data really relates to non, non-permanent resident, non-US citizens who are interacting with various facets of DHS, uh, as they come here. So it's data on specific encounters, encounters they have with Border Patrol, criminal justice system, and so on. But does it extend to things like last known address? Is that the kind of data in?
When it comes, so what was the address they reported in the border account? What was the address they reported when they applied for benefits?
Right. So then that becomes the kind of data that if I'm doing a raid, they end up using, right, to say, okay, we have this person here and so on. So, so it is sort of effectuating raids through that mechanism.
To what extent does it go beyond that? Like, is there integration with local law enforcement, for instance, with any of this stuff?
Okay. But I, I mean, what is the, I think the real question, just zooming out a little bit. We're working with agencies that have lawful authorities. So what, what data are they collecting? That's the data we're helping them integrate. So if there's a lawful authority for them to collect the data, you know, I don't, I don't think local law enforcement is part of this. I'm, I'm not saying I have an objection to them being part of this. I'm, I'm just saying that.
No, I'm just, I'm just trying to get into, I mean, I think the, the sort of the core fear that people have about, well, there's sort of two layers of fears that I think people have with this kind of domestic data collection, and that are distinct from people who just oppose deportation. One layer, which, as far as I can tell, is sort of overstated or wrong, is that Palantir is, you know, again, going out and devising no data.
Right, you're not devising, but you're not devising systems to acquire new, new forms of data for the government, right?
But you are creating systems where the government has access to data, and I, I think this is important, on a scale that no government has ever had access to before, right? Like one of the unique features of 21st century America, digital age America, is that, you know, it's just so much easier to track and observe people in all kinds of different ways. And corporate America does that even more than the government does, right? But.
I think so. I don't think that's the fundamental inflection, though. I think it's really that, um, the number of decisions you're trying to make are growing much more quickly than your headcount is growing. So let's just say in the ICE case, uh, you would have had 20, roughly on the order of 20 systems that you would have to manually go search across. You know, how many times.
If you're trying to find the address, the address of someone who you want to deport.
Right. Um, do you think you're going through all 20 systems? You know, maybe you stop at 10, maybe forgot to search some systems. So the, the entropy is just that you're not actually able to do the work. You know, you're fighting your tooling the whole time. And this sort of leads to the lack of legitimacy of the institution. It's like you're not able to do the thing that you're actually supposed to do. Of course, the complexity of this is is growing as you're asked to do more and more. You know, it's like maybe I was doing 10 things before and I could do it with the people I had. Now I have to do a thousand things. That means I clearly don't get to all thousand things. And the things I get to, I'm not doing them. Well, it's like, okay, well, why are we using the human to search 20 different systems? That's clearly a problem that technology can do. How do we use the human to do things that are uniquely human, judge that require human judgment?
Right. I guess I'm just saying that I think for at least some people, and maybe I'm among them, there's a kind of relief in the idea that all of this data has multiplied so quickly that people can't, you know, that sort of authorities can't quite keep up with it. Right? So if we walk outside, if we walk outside of this room and enter, you know, Midtown Manhattan, we are under constant surveillance. Now, it's not all, it's not government surveillance, right? But there's a kind of relaxation that you can feel where you're like, okay, but all of this surveillance is, you know, distributed across so many different public and private entities, and unless I am literally a terrorist, the odds that people are going to be constantly watching and scrutinizing me are very low. But then the fear becomes, you know, well, if we have this incredible way to make it all more and more and more efficient, then maybe privacy does start to disappear. I don't know. What do you, what do you make of that?
There's, there's two thoughts there. One is, well, you know, are you saying that you feel safer because the institutions that are supposed to protect you are structurally incompetent? And that's the part where I feel like.
The answer might be yes, sometimes. Right.
Yeah. And then a consequence of that, which, you know, I think a democracy can decide, a consequence of that is that they also can't do their job. They can't protect you from the things that they're supposed to protect you from. So, I, I'd offer another solution to this, which is they should be really good at doing what they're doing, and we should have a strong ability to oversee that they're not doing things that they're not supposed to be doing. That's exactly what we designed Palantir to do. So, this wasn't our system, but I think it's an illustrative example because it was high-profile at the time. If you go back in time, um, there were government employees who looked up the passports of Hillary Clinton and Barack Obama, and they got caught. How did they get caught? The system they're using generates audit logs. People were reviewing the audit log and, you know, did you have a permissible use to look this up? The answer was clearly no, and they were, they were quickly caught and terminated.
Right.
You know, and so you have to both help, uh, the people who are protecting us, and you have to empower the people who are watching the watchers, so they also have asymmetrically strong technology to ensure that there are no abuses.
So who is watching the watchers? It's like, so let's, we can stick with immigration enforcement, right? So presumably, there's data and information that is protected by privacy that, that, you know, that ICE is not supposed to be accessing, right?
So the, the first line of defense is on the front end, like, do you even have authority to have this data? So, you know, the lawyers on the front end, you have data use agreements that they have with, uh, with their own inter-agency data use agreements. The authorities they collected the data to begin with. And then the second line of defense is after they have the data, is are you using it for permitted use? And that's usually, you know, I don't know for every agency, this is, look, I'm a, I'm a builder, so you consult a policy person, but something like the Office of the Inspector General, um, other law enforcement agencies that are responsible for this, sometimes depending on the context, it can actually be counterintelligence. It's the same, the same infrastructure that you would use to make sure that people aren't abusing the data is how you would also make sure that, you know, you don't have a spy in your organization.
Right. In the end, though, it is up to the people running those institutions to decide.
Whether, you know, whether to basically track abuses at all. Right? So, and it might be at different levels and different agencies, but like, presumably in the Department of Homeland Security, Kirstjen Nielsen would have the authority to, you know, track or not track abuses to some degree, right?
I mean, yeah, I think that's a little bit of a nihilistic view. I mean, there's statutory requirements on this. So I think there's a fundamental question if, you know, I, I understand the perspective. I don't agree with it, that hey, we can't trust these institutions at all. It's like, there are rules, they follow the rules, they're also humans, they're infallible, but I don't think people are like willy-nilly deciding whether they want to do this or not.
No, I don't think so either. I'm just trying to frame what I think are the most commonplace criticisms and critiques and worries about how this technology is used in the context of, you know, the second Trump administration. You know, I, I think from the point of view of someone, someone who is a skeptic of Palantir might say, well, in the end, you are, you are putting a lot of power into the hands of people that, you know, this person who thinks Donald Trump is a threat to the republic, right, doesn't trust, right? So there is, there is, there is just a way in which you, independent of what you think about Trump himself, when Palantir goes to work for a government, US government, any other government, you are putting yourself in the position of trusting that government with this very impressive technology that you've built, right?
Yeah, I, I think that's right. That's you have to pick your customers.
Right. So, can I just ask then about that, about that process? Because I think it's sort of interesting, like, how, how do you calibrate that kind of decision-making? Like, so just recently, um, the UK, United Kingdom rolled out a plan to have a new digital ID, and [snorts] I believe the head of Palantir in the UK said, we don't think there's enough privacy safeguards here, and we're not going to, we're not going to cooperate with the government on this. So that's like an exact example of making this kind of, sort of practical, but also moral decision. What kind of thinking goes into a decision like that?
Well, I'm probably not the best person to comment as the CTO, but I think I'd have to go back to the quote. I think what, um, Louis Mosley said was that this is an issue for the ballot box.
Louis Mosley is the head of Palantir in the UK.
That's right.
Right.
So, um, look, there's a lot of discursive interaction both internally, uh, at Palantir. There's an obvious threshold question of, is this a legal use? Is this a government that you trust and work with? Right. Uh, and then beyond that, it's, it's really a question of like, are the authorities there? What's the potential for misuse?
Um, are the protections there in our software against those potential abuses? U and is the work actually going to lead to something meaningful?
See, that's really interesting though, right? That that he said it's a matter for the ballot box. And again, I know you're not, you're not running that decision, but it's still, as an outsider to the work your company does, it seems like he, or the larger corporate enterprise, made a judgment that basically a political judgment that the digital ID issue had not been adequately litigated by British democracy for you to cooperate with it. Isn't that kind of a fascinating, it's a fascinating decision. I, I mean, I.
You talked to Louis about that. I would, I would give you the other side of that, which is I think a lot of the policies, a lot, a lot of things that people are struggling with right now, I think in the US, were voted on at the ballot box. You know, what ICE is doing, right, was voted on at the ballot box. Um, so.
So, right. No, so you, so Palantir, again, [snorts] not that you speak for everyone at Palantir, but Palantir is more comfortable with mass deportations in the context of a world where Donald Trump campaigned on mass deportations and won, than it might be in a world where he had not mentioned deportations at all and then asked Palantir to design mass deportation embedding software.
What I'm saying is like, the people voted on this, right? You know, that seems like a functioning democracy there. And of course, there's going to continue to be discursive interaction and disagreement. That's our, that's the beauty of our political process, but it's an important distinction.
Yeah. How do, how do you think about that with foreign clients? Like, obviously, Palantir does work for the state of Israel, that is again, obviously a source of controversy. But so you're making a judgment that Israel is an appropriate partner. Yes.
Do you work with dictators, with non-democratic governments?
We, we do work with the US and our allies. Really, the, the nature of the work depends particularly, I think, in the coalitions that are fighting, uh, terrorism, the coalitions that are pushing back on, um, the great power competition that we're in.
With the Big Four?
But so would you be un- if the Kingdom of Saudi Arabia called you up and said, we just want to build a digital ID system, you would be less likely to do that work because they're not a democratic government? Would that be fair?
I would, I would leave that up to, to Alex to figure out. It depends on their uses. I'm not, I'm not sure that's exactly fair. I mean, um, is it leading to a more efficient electronic government? You know, what's the, what's the purpose of it is the question. There's a difference of which agencies doing it and what is their mandate?
What data do they have? How are they trying to, um, what are the workflows you're trying to enable? But from the beginning, right, just again, in the name from Lord of the Rings, right, you have this idea that you are sort of, you're building a technology that in the story the name originates from is used for good and for evil.
I'm just interested in the extent to which that kind of judgment is more than most companies woven into the kind of governmental work that you're willing to do.
I mean, we, we, we, we pick our partners very carefully. Uh, we want to work on work that's important to the West, and, um, that includes the commercial work we do in re-industrializing the country.