AILSA CHANG, HOST:
Anthropic's newest AI models have been suspended by the federal government over national security concerns. This happened on Friday evening, and officials from the tech company and the Trump administration are still debating a solution. It's just another example of conflict between the Trump administration and Anthropic, a company that has advocated for more guardrails around the technology. NPR White House correspondent Deepa Shivaram is here with more. Hi, Deepa.
DEEPA SHIVARAM, BYLINE: Hello.
CHANG: Hello. OK. So the government issued this suspension of the new models from Anthropic on Friday. Well, what's happened since then?
SHIVARAM: Yeah. So those Anthropic models are Fable 5 and Mythos 5, and they were suspended because of national security concerns. The threat was that the models could be jail broken. In other words, the security mechanisms built into the AI models could be undone. Now, the suspension from the government was for any foreign users, so anyone not a U.S. citizen. And that even included some Anthropic employees. But in order to comply with the directive, Anthropic said they had to shut down their models for everyone. So since that's happened, there's been, as you can imagine, a lot of confusion and disruption and definitely still questions about the mechanisms and the legal authority the administration has to suspend this technology.
The White House, on their end, has been mostly quiet about the decision, but they did say that officials from the Commerce Department met with Anthropic in Washington yesterday. Anthropic said in a statement that, quote, "both parties are working quickly to get this resolved."
CHANG: Wait. So shutting down access so abruptly like that, like, how does that affect the Trump administration's goals for AI because, I mean, they've historically been more interested in innovation rather than regulations around safety, right?
SHIVARAM: Yeah. I mean, there's still a lot unfolding here when it comes to what those specific national security concerns were, who they involve. But in terms of immediate impact, there's some big concerns here about what kind of precedent this sets, Ailsa. I spoke with Ifeoma Ajunwa. She's a law professor who focuses on AI ethics at Emory University, and she said this suspension could have what she calls a chilling effect on the U.S.'s lead on AI around the world because it may not be a reliable place for AI businesses or startups.
IFEOMA AJUNWA: If a company feels like it's not going to be able to globally expand and that the U.S. is an unwelcoming business home, then, you know, that company is going to go elsewhere.
SHIVARAM: And this is kind of part of this recent trend, almost, from the Trump administration of how they're sort of undercutting themselves on AI. Trump's goal with AI has been to win the global AI race, to beat China, to innovate, to innovate, to innovate. But many of his decisions, whether it's, you know, cutting science research funding or limiting foreign student visas or this suspension, a lot of experts say it counteracts some of those goals. And it also throws into question what kind of relationship the U.S. wants with its allies when it comes to sharing the benefits of AI, right? This is a hugely impactful technology. It's changing our lives every day, so what happens when there are limits on how much it's shared and who we share it with?
CHANG: Yeah. Well, speaking of allies, Trump is at the G7 now in France. Has this issue come up in meetings over there? Like, what are other world leaders saying about all this?
SHIVARAM: Yeah. There's definitely a lot of global leaders who are very tapped in, very much paying attention. Canada's Prime Minister Mark Carney said that this was a warning for what overreliance on certain models can look like. I think, looking forward, the EU is something to watch. They've been pretty strict on their AI laws, which the U.S. has been critical of, but maybe they won't want to rely on U.S. technology as much anymore and change some of those regulations. So that's something to look out for.
CHANG: That is NPR's Deepa Shivaram. Thank you so much, Deepa.
SHIVARAM: 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.
