LEILA FADEL, HOST:
911 dispatchers face a new challenge. AI-generated voices or background sounds are making hoax calls sound more realistic. That makes it harder to figure out what's real. NPR's Jaclyn Diaz reports.
JACLYN DIAZ, BYLINE: In August, a caller made a bomb threat to Wheatland Union High School in Yuba County, California. Police responded overnight, and classes were canceled. There was no bomb, and no one was hurt. Local station FOX40 reported that an AI-generated voice was used.
(SOUNDBITE OF ARCHIVED RECORDING)
UNIDENTIFIED REPORTER: The superintendent says the Wheatland police received an anonymous bomb threat overnight, where an AI-generated voice was used, demanding $15,000 in a container box.
DIAZ: Authorities said it was part of a series of hoax calls, also known as swatting, that were targeting schools around the country over the summer. Swatting calls are usually made to 911 or police departments. Callers report a bomb threat or that they're under attack of some kind. The goal is to draw a big response from police to a home, business or school. It causes a lot of chaos and sometimes even worse.
MEL MAIER: People have been killed and injured in these swatting attacks.
DIAZ: That's Mel Maier, the CEO and executive director of APCO International, which is the Association of Public Safety Communication Officials. Its members include 911 dispatchers. APCO and the Public Safety Threat Alliance, an information-sharing and cybersecurity group, looked into the prevalence of swatting to 911 call centers. They published their findings in June, based on more than 760 responses. About 14% of dispatchers said they were targeted using AI tools. Maier says he thinks that number is going to keep going up with increasingly advanced tools.
MAIER: Those voices used to be pretty easy to recognize, but now they're natural-sounding. They can use words like, uh, or maybe, um, let me think about it. And it sounds very natural.
DIAZ: Swatters are also using AI to make it sound like there's gunfire or screaming in the background. Brandon May is an assistant professor of forensic psychology at the Florida Institute of Technology. He says tools to create such sounds are cheap and readily available from companies like ElevenLabs, an artificial intelligence company creating synthetic voice software.
BRANDON MAY: I'm able to clone my voice on ElevenLabs for less than $20, and I can change the intonations. I can change how deep or high-pitched my voice is. I can change the pauses.
DIAZ: He says it's even hard for him to detect AI fakes. And that's really concerning, but even more so for dispatchers who need to make decisions fast.
MAY: You couple that in a high-stakes, high-consequence 60- to 90-second call - it is going to be nearly impossible to detect.
DIAZ: So groups like APCO are rolling out training for dispatchers so they're better able to determine which calls are legitimate. What does that look like? Here's Maier again.
MAIER: That's to create criteria-based caller interrogations where specific questions are asked with such granular level that an AI cannot scoop that information from a public database or simply make guesses.
DIAZ: But even if dispatchers strongly believe a call is fake, Maier and May say they will always react as if the threat is real and will send resources to any call for help.
Jaclyn Diaz, NPR News.
(SOUNDBITE OF MUSIC) 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.
