A MARTÍNEZ, HOST:
The artificial intelligence company Anthropic is saying no to the Pentagon. The CEO of Anthropic, Dario Amodei, says his firm won't agree to let its technology be used for mass domestic surveillance and fully autonomous weapons. Now, the showdown between the two is one of the first between this new industry and government interests. For more on this, we turn to reporter Maria Curi at Axios. So let's start off, Maria, with how the military uses Anthropic now.
MARIA CURI: The military uses Anthropic for research and development and going through documents. That's mostly the information that we have right now. You know, the Department of Defense won't tell us exactly how it is using artificial intelligence for all of its operations, especially in the types of operations that Anthropic is most important for here, which is classified settings. But we do know, for example, that the technology was used in the Maduro raid.
MARTÍNEZ: OK. And these two categories - these categories that the military uses Anthropic, those are red lines for the company?
CURI: Yes. So essentially, Anthropic has set these two red lines. One is around mass surveillance, and the other is around using Claude, its model for autonomous weapons. That's weapons without a human in the loop. And, again, we don't know exactly how the Pentagon uses Anthropic for classified settings, but we know that Anthropic's uses policies were not violated during the Maduro raid, and so those two capabilities were not necessary.
MARTÍNEZ: All right. So last night, Pentagon Chief Technology Officer Emil Michael said on social media that Anthropic CEO, quote, "has a God-complex. He wants nothing more than to try to personally control the U.S. military and is OK putting our nation's safety at risk." What else is the Pentagon saying about this?
CURI: So here's the Pentagon's position in all of this. They essentially want to be able to use Anthropic's technology, and not just Anthropic's technology, but Google's technology, OpenAI's technology, xAI's technology for all lawful purposes, for classified and unclassified settings. The tricky thing about all lawful purposes is that the law today doesn't necessarily contemplate artificial intelligence. And so right now, it is legal for the government to, for example, collect all public data on Americans. That includes social media posts, whether you have a concealed carry permit, if you've attended protests or rallies. And you can imagine how artificial intelligence, injected into those legal data collection capabilities would make it much easier to continuously and quickly collect that data, target specific individuals and surveil them. That's where Anthropic is stepping in and saying, you know, all lawful purposes does not necessarily work for us because the law has not caught up.
MARTÍNEZ: It seems funny that the law is always behind technology in almost everything in the last 20, 30 years. So OK, the government - the U.S. government - and Anthropic seem to have a lot to lose here, but who loses more if they somehow wind up dumping each other?
CURI: You know, both sides, like you mentioned, really have a lot to lose. This is a $200 million contract that's on the line for Anthropic. Now, this company is worth $34 billion, a drop in the bucket, but they're being threatened with, you know, being designated a supply-chain risk, which means that any company that does business with the Pentagon can't do business with Anthropic, and that would be a big low. But the Pentagon right now, it's really hard for them to untangle Anthropic. And so I would say that as of now, the Pentagon has more to lose.
MARTÍNEZ: All right. Maria Curi, tech policy reporter at Axios. Maria, thanks.
CURI: Thank you. Transcript provided by NPR, Copyright NPR.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.
