ARI SHAPIRO, HOST:
Two days before New Hampshire's primary last month, thousands of people in the state received what seems to have been a robocall generated by artificial intelligence. A voice that sounded a lot like Joe Biden told people not to vote in the primary election. Authorities immediately announced an investigation, and now they say they know the source of the calls. The incident made the warnings that have been coming from many election experts suddenly very real. NPR's Miles Parks covers voting for us, and he's here in the studio. Hey, Miles.
MILES PARKS, BYLINE: Hey, Ari.
SHAPIRO: So who was responsible for that New Hampshire robocall?
PARKS: So yesterday, the New Hampshire attorney general announced that they had traced the calls back to a Texas-based company called Life Corporation and to a man named Walter Monk. They are not pursuing charges at this point - they didn't announce that yesterday - but they say they have sent out a cease-and-desist order as well as subpoenas for records and that their investigation is ongoing. Our colleagues at NHPR have reached out to all the parties involved here, have not heard anything back yet, but notably, we also learned more details about how many people actually received these calls. They estimate that between 5,000 and 25,000 people received these calls ahead of the primary, which, obviously, like you said, makes this a very real concern looking ahead to November.
SHAPIRO: That's huge. NHPR, we'll say, is the local member station, New Hampshire Public Radio.
PARKS: Yep.
SHAPIRO: The night of the primary, you were telling me how worried experts were about the potential for AI to spread disinformation in this year's elections. How much did this robocall drive that home?
PARKS: I mean, it is crazy how much this mimicking technology has gotten better really, really quickly. And the other thing I've heard a lot is something many experts have said - is that AI gives the potential for tracking a lot better and lets bad actors potentially target this sort of information as well. You think about an average person being able to send a message like this out to, say, 10,000 swing voters in Arizona. I talked about that with Joe Kiniry, who's an election expert who works for a company that works on verifiable election systems.
JOE KINIRY: The power afforded by new technologies that can be used by adversaries - it's going to be awful. I don't think we can do science-fiction writing right now that's going to approach some of the things we're going to see over the next year.
PARKS: You know - I know. Hyperdetailed voter files have been a part of campaigns for a while, but Kiniry said he cannot imagine all the different use cases when you start to add in AI.
SHAPIRO: What kinds of tools do election officials have to counter this?
PARKS: I mean...
SHAPIRO: How are they responding?
PARKS: ...I will say some election officials kind of pumped the brakes on some of the concern. There's a wide spectrum of just how bad this could be this year. The arm of the Department of Homeland Security that works to secure elections put out a report last month that said AI will likely not introduce new risks but could amplify the risks that already exist, so I think a lot of election officials are looking at this. Like, cybersecurity and disinformation were two of our biggest concerns in 2020. We've been working on those, and AI just presents the potential for those problems to get worse. I think one of the things that I keep hearing from election officials is just begging voters to go to trusted sources of information when they're thinking about voting and thinking about elections.
SHAPIRO: Yeah. What advice do you have for voters to make sure they don't get duped by one of these?
PARKS: It sounds really basic, Ari, but getting comfortable going to your county or local officials' website - usually, that's a website that ends in .gov - or even calling the office. I think election officials are kind of resigned at this point to the threat environment - that there is going to be bad information floating around there. And building resilient voters - voters who, when they watch a TikTok, when they get a robocall that sounds like Joe Biden telling them not to vote, their impulse is not going to be to listen to that but instead to go check it at a county website, for instance, because this stuff is just getting easier and easier to make.
SHAPIRO: Well, I hope this is the last time you're here talking about a deepfake in this election, but I fear it might not be.
PARKS: I doubt it will be.
SHAPIRO: NPR's Miles Parks. Thank you.
PARKS: Thanks, Ari. 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.
