Full interview: Anthropic CEO Dario Amodei on Pentagon feud

That's correct. Yeah.

Yeah. So, you know, we should maybe back up a bit for a little bit of context. So, Anthropic actually has been the most lean forward of all the AI companies in working with the US government and working with the US military. We were the first company to put our models on the classified cloud. We were the first company to make custom models for national security purposes. We're deployed across the intelligence community and the military for applications like cyber, combat support operations, various things like this. And the reason we've done this is I believe that we have to defend our country. I believe we have to defend our country from autocratic adversaries like China and like Russia. And so we've been very lean forward. We have a substantial public sector team. But I have always believed that as we defend ourselves against our autocratic adversaries, we have to do so in ways that defend our democratic values and preserve our democratic values. And so we have said to the Department of War that we are okay with basically 98 or 99% of the use cases they want to do except for two that we're concerned about. One is domestic mass surveillance. There we're worried that things may become possible with AI that weren't possible before. An example of this is something like taking data collected by private firms, having it bought by the government, and analyzing it in mass via AI. That actually isn't illegal. It was just never useful before the era of AI. So there's this way in which domestic mass surveillance is getting ahead of the law. The technology is advancing so fast that it's out of step with the law. That's case number one. Case number two is fully autonomous weapons. This is not the partially autonomous weapons that are used in Ukraine or could potentially be used in Taiwan today. This is the idea of making weapons that fire without any human involvement. Now even those, I think that our adversaries may at some point have them. So perhaps they may at some point be needed for the defense of democracy. But we have some concerns about them. First, the AI systems of today are nowhere near reliable enough to make fully autonomous weapons. Anyone who's worked with AI models understands that there's a basic unpredictability to them that in a purely technical way we have not solved. And there's an oversight question, too. If you have a large army of drones or robots that can operate without any human oversight, where there aren't human soldiers to make the decisions about who to target, who to shoot at, that presents concerns and we need to have a conversation about how that's overseen and we haven't had that conversation yet. And so we feel strongly that those two use cases should not be allowed.

So, there were several stages of this, all done quickly and determined by the very limited 3-day window that they gave us, right? They gave us an ultimatum to agree to their terms in 3 days or be designated a supply chain risk or defense production act. I guess we'll get to that later. But during that time, there were a few back and forths. At one point they sent us language that appeared on the surface to meet our terms, but it had all kinds of language like if the Pentagon deems it appropriate or to do anything in line with laws. So it didn't actually concede in any meaningful way. And there were further steps of it that also did not concede in any meaningful way. We have wanted to strike a deal since the beginning. If you want to get a sense of the Pentagon's position, the Pentagon spokesman Sean Parnell the day before tweeted, he reiterated their position. We only allow all lawful use and this was the same as when they sent us their terms. So they have not conceded and they have not in any way agreed to our exceptions in any meaningful way.

So, in the statement we issued yesterday and also in the one we issued today, we said that we were willing even if the Department of War or even if the Trump administration takes these unprecedented measures against us, this kind of supply chain designation that's normally used against foreign adversaries, we have said that even if they take these extreme actions, we'll do everything we can to support the Department of War to provide its technology for as long as it takes to offboard us and onboard a competitor who's willing to do these things that we are not willing to do.

Yeah. So, we have offered continuity. We're actually deeply concerned about this. We're deeply concerned about the interruption of service, which is exactly what's happening when we're designated a supply chain risk, right? When we're designated a supply chain risk, they say you have to be off all of our systems. And I've talked to people on the ground, uniform military officers, who say this is essential. Not having this will set us back six months, 12 months, maybe longer. And so that's why we've tried so hard to try to get a deal. But again, the 3-day ultimatum, the risk to designate us a supply chain. The whole timeline has been driven by the Department of War, not by us. We are trying to provide continuity. We are trying to provide the services. We are trying to reach a deal here.

Yeah. I would say a couple things. In the short run, it means and it's up to the Department of War. We're still trying to reach a deal with them. We're still trying to talk to them.

We've received various communications. We haven't seen anything that satisfies our concerns. But I mean that just in the broad sense that we are still interested in working with them as long as it is in line with our red lines.

You know, look, an agreement requires both sides. We for our side are willing to serve the national security of this country. We are willing to provide our models to all branches of the government, including the Department of War, the intelligence community, the more civilian branches of the government under the terms that we've provided, under our red lines. We are always willing to do that, right? We don't take offense here. The reason we're providing our technology in this way is that we want to support the national security of the United States. We're not doing it for the sake of Pentagon officials. We're not doing it for the sake of a particular administration. We're doing it because it's good for the national security of the United States, and we're going to continue to do that.

So, first I would say, and I think this is an important point, no one on the ground has actually to our knowledge run into the limits of any of these exceptions. These are 1% of use cases and ones that we have seen no evidence on the ground have been done. Now, again, I can't say what their plans are that we don't know. But we have no evidence that these use cases have actually run into trouble. We've spread across the Department of War and other parts of the government without running into any of these problems. Now, in terms of these one or two narrow exceptions, I actually agree that in the long run, we need to have a democratic conversation. In the long run, I actually do believe that it is Congress's job. If for example there are possibilities with domestic mass surveillance, government buying of bulk data that has been produced on Americans, locations, personal information, political affiliation to build profiles and it's now possible to analyze that with AI. The fact that that's legal, that seems like the judicial interpretation of the Fourth Amendment has not caught up or the laws passed by Congress have not caught up. So in the long run we think Congress should catch up with where the technology is going but Congress is not the fastest moving body in the world and for right now we are the ones who see this technology on the front line. I would expect the Department of War to be thoughtful about these issues, to proactively think about these issues and so I would have expected them not to have any concern and for us to have a conversation but I think in the absence of that, we need to look at the technology, we need to look at what it's capable of in terms of reliability and we need to look at the ways in which it's getting ahead of the law and in which it's escaping the intent of the law. Those are some very narrow areas, but I think that they're important. These are things that are fundamental to Americans, right? The right not to be spied on by the government, right? The right for our military officers to make decisions about war themselves and not turn it over completely to a machine. These are fundamental principles.

Well, I would give two answers to that. One, we are a private company, right? We can choose to sell or not sell whatever we want. There are other providers. If the DoD, the government, doesn't like the services we provide or the way that we make them, they can use another contractor. This would have been the normal way to handle this, right? Just to say I would have disagreed, but I would have respected them if they said DoD, we don't want to work with Anthropic. Our principles are not aligned with yours. We're going to go with one of the other models. But they've both extended that to parts of the government beyond the DoD and tried to punitively revoke our contracts beyond DoD and they've done this supply chain designation thing which basically says that if you're another private company who has military contracts, you can't use Anthropic in a way that touches those military contracts. So they're reaching in to the behavior of private enterprise. And it's very hard to interpret this in any way other than punitive. To our knowledge, the supply chain designation has never been applied to an American company. It has only been applied to adversaries like Kaspersky Labs which is a Russian cybersecurity company that is suspected of ties to the Russian government, Chinese chip suppliers. Being lumped in with them, it feels very punitive and inappropriate given the amount that we've done for US national security.

We don't. Look, one of the things about a free market and free enterprise is different folks can provide different products under different principles. Remember, this isn't just about terms of use. This isn't just about this is what our model is legally allowed to do. Our model has a personality. It's capable of certain things. It's able to do certain things reliably. It's able to not do certain things reliably. And I think we are a good judge of what our models can do reliably and what they cannot do reliably. And I think we do have a good view into how the technology again is getting ahead of the law. But I would say again, I actually agree with you that this is not tenable in the long term. I don't think the right long-term solution is for a private company and the Pentagon to argue about this. I think Congress needs to act here and we are thinking about that. We are thinking about what Congress could do to impose some of these guardrails that don't hinder our ability to defeat our adversaries, but that allow us to defeat our adversaries in a way that's in line with the values of our country. But as you know, Congress doesn't move fast. So, I think in the meantime, we do need to draw a line in the sand.

Again, I would say two ways that it's different. One, I would point again to the newness of the technology. Right? When a technology is well-established then I mean there are lots of technical things about aircrafts but I think a general has a pretty good understanding of how an aircraft works, like aircrafts have been around for a long time.

Sure, but not at the pace that we see with AI. AI is moving so fast. I've talked often about how AI is on an exponential trend. The models, the amount of computation that goes into the models doubles every four months. We have never seen anything like this pace of innovation.

I think there's only catching up once, right? So the pace of the technology is fast, the issues that arise are few but very important. Again, we only have two of these: domestic mass surveillance, fully autonomous weapons. We need to have a conversation with Congress to help them understand some of the risks associated with it. Again, this is the most American thing in the world. No one wants to be spied on by the US government.