JOEL ROSE, HOST:
Sometimes journalists who cover a beat have an aha moment, something that marks a clear before and after. That was the case for Brett Neely, one of NPR's editors who oversees our coverage of artificial intelligence. It was about four years ago when he was editing our mis- and disinformation coverage of conspiracy theories.
BRETT NEELY, BYLINE: The first OpenAI, like, image generation models started appearing.
ROSE: He could tell this had the potential to cause chaos.
NEELY: You could just start creating images of things and spread them online via social media and really confuse people about what actually happened.
ROSE: Over the next few years, deepfakes became more and more common. And then in 2025, OpenAI released a video generator tool called Sora 2.
NEELY: One of the first things I did when I got my hands on Sora was to make a grainy security-style video of people stuffing ballots into a ballot dropbox because that was, like, exactly the scenario I feared this stuff could be used for, and it just generated it like that.
ROSE: Sora was later discontinued, but Brett's work - editing and assigning stories for our coverage of AI - has only expanded, and that was before the resignation of Jacob Coxon, a researcher at the AI company Anthropic, who resigned over safety concerns last month, focusing the world's attention on AI unlike any other story before. So when I sat down with Brett for this week's Reporter's Notebook, I started by asking him how that story informed NPR's reporting on AI.
NEELY: I think that it has changed the pace, and I think that there is a lot more interest from our audience and, you know, from our news magazines and elsewhere for this kind of coverage. I don't necessarily think it has changed how we cover things. But, you know, we are sort of going back, and we have done some conversations as well that have just been a little bit more explanatory in nature. You know, I mean, I remember the week after the Coxon story broke, you know, editing just a basic conversation about why are there fears for humanity in this and really trying to untangle, like, what is hyperbole and what we should actually be concerned about.
ROSE: Can you point to a recent story that your team aired that you think encapsulates, I don't know, NPR's approach to reporting on AI?
NEELY: It would be hard for me to point to one. I feel like we've done a lot on AI lately.
ROSE: Fair.
NEELY: We've been doing some - just some very sort of explanatory reporting. For example, you know, out of this latest surge of interest, you know, in AI safety, both Anthropic and Open AI have been talking about how they want to have these third-party evaluators sit in and evaluate their products before they're released. And so Huo Jingnan, who's one of the reporters on our team, you know, did a very, like, good and careful story kind of looking at, you know, who these evaluators are. And there are a couple of big issues with them, one of which is, the evaluators say they're not getting enough time to evaluate these model.
(SOUNDBITE OF ARCHIVED NPR CONTENT)
HUO JINGNAN: All the researchers I talked to say that evaluation results don't do a very good job of predicting model behavior. They can't predict all the ways people will interact with AI models, and they can't predict how AI models will behave in situations that evaluations didn't test for.
NEELY: The other thing that we are trying to do is also be very clear that, like, it's not just these large, abstract, theoretical risks that people are concerned about, but, like, we just need to be talking about very specific things that are happening today that actually, you know, many people feel like are causing harm. We shouldn't just be talking about whether or not we're building Skynet and going to launch an army of Terminators on us. Like, I think that we should be talking about what's happening here and now.
ROSE: Although we probably should not build Skynet.
NEELY: We should not build Skynet either.
ROSE: To be clear, NPR just this week announced a grant from the Tarbell Center for AI Journalism to support this coverage that we've been talking about. And I know you can't speak officially for the fundraising side of the organization on this, but listeners might have questions about how this is going to work. What can you say about how it'll work?
NEELY: I mean, I would say that NPR regularly gets grants from foundations to cover certain subject areas. One of the conditions that NPR always puts is full editorial independence, and that is certainly also the foundation's policy. So, you know, they're happy to support coverage. They don't tell us what to do. We don't show them our work or anything like that. We go and do what we think is important, and we don't let the funders sort of dictate what we're going to cover.
ROSE: When you and your team are talking about what to cover, what are the big questions that are driving your reporting right now?
NEELY: I feel like this is now a beat that can't really separate the sort of politics and policy from the business side. And so I think increasingly, there's just going to be a lot about how much of a political story this is, whether it's just in terms of voters, but also what future regulation is going to look like. You know, I think that there's a lot of reporting to do from inside the companies as well and from, you know, the safety cultures that are present.
And, you know, I mean, we just keep seeing this contradiction of almost everyone in these companies saying, from the CEOs on down, sort of talks about what a powerful scary technology this is, and it really needs to be regulated. But, you know, then they keep shipping new releases, like, almost weekly and also feel like they can't stop. And I think really getting to the heart of that dynamic and that contradiction is just going to be a really important theme.
ROSE: Do you ever look back at your story list and be like, oh, we're kind of the doomer desk this week?
NEELY: (Laughter) I mean, you know, it's sometimes - you know, it's sometimes harder because the benefits haven't totally materialized, right? We are seeing disruption now. But the benefits of, like, you know, how does this sort of reorder business and things like that - like, those are still a little bit theoretical at this point.
ROSE: Is there a story that you wish you had covered more, that you had put more resources on? And why do you think it did not get the treatment that you think it should have?
NEELY: I think when the first trickles of information about the Hugging Face incident - that is where OpenAI disclosed that some of its agents, which are essentially autonomous programs, had hacked this other AI company, Hugging Face, and also hacked OpenAI. And that kind of led to the string of revelations about all the various websites that these agents had disrupted or at least attempted to or just, you know, done things that they weren't actually programmed to do.
I don't think we understood the gravity of it immediately in the moment, and it took us a few days to really respond, I think, in the way that I felt good with. But that was also because, like, the information was kind of coming out in dribs and drabs, and it was a little bit hard to understand.
ROSE: What is your personal relationship like with AI right now, if you don't mind my asking? Like, do you feel like you have to be using it regularly in order to cover it well?
NEELY: I think it's actually quite complicated just because, you know, I'm never going to edit a story with AI. I'm not going to use it for writing in any way whatsoever. So I would say that I'm probably using it less than listeners might expect.
I'm - you'll probably get a lot of flame mail for this. It's not that I don't use some of the different chatbots from time to time, but I don't have them integrated into my workflow because I don't know a way to do that that I feel like would be sort of true to our promise of delivering news made by humans for humans. So in that sense, I'm maybe not as much of a power user as some people.
But also there's just the level on which, like, this is actually already embedded into our lives in ways that we're not even, you know, fully conscious of. And it's like, you know, every time you do a Google search result, there's an AI summary, and those AI summaries are quite inescapable. And even if you say that you don't want to look at the AI summary, it's hard to avoid it.
ROSE: I've been speaking with Brett Neely. He is the supervising editor of NPR's team covering power and influence. Brett, thank you very much. Thanks for your time.
NEELY: Thanks, Joel. 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.
