An August bomb threat to Wheatland Union High School in Yuba County, California, was a hoax. Police responded overnight, classes were canceled, and no one was hurt. What set the call apart, NPR reported, was that authorities said it was made with an AI-generated voice, part of a wave of swatting hoaxes that 911 officials say is getting harder to recognize.
Swatting is the practice of reporting a bomb threat, mass shooting or other life-threatening emergency in order to draw a massive police response, often to harass or retaliate against a target, according to the FBI as cited by NPR. People have been killed and injured in swatting attacks, said Mel Maier, CEO of APCO International, which represents public safety communications officials. Dispatchers used to be able to pick out many hoax callers. That is changing.
"Those voices used to be pretty easy to recognize, but now they're natural sounding. They can insert pauses. They can use words like, 'uh,' or 'maybe, um, let me think about it.' And it sounds very natural," Maier told NPR.
Cheap clones, and calls that skip 911
APCO and the Public Safety Threat Alliance published findings in June based on 766 responses from public safety communications professionals. About 14% of dispatchers said they had been targeted using AI tools. Maier said he expects that share to rise. Brandon May, an assistant professor of forensic psychology at the Florida Institute of Technology, told NPR that tools from companies such as ElevenLabs are cheap and easy to use. "I'm able to clone my voice on ElevenLabs for less than $20, and I can change the intonations," he said. "It is no longer simple to just look at an image or hear a video or hear a voice and go, 'That is AI.'"
April Heinze, vice president and chief of 911 operations at NENA: The 9-1-1 Association, said most swatters do not dial 911 directly, because a true 911 call provides location, a callback number and the name of the phone's owner, and "it's really impossible to spoof your location on 911." Calls are more often placed to police departments or courthouses and then transferred into 911 centers, which she said makes tracing them effectively impossible.
APCO and NENA are developing new training but declined to describe the specifics, saying more public detail would only help people train AI to fool dispatchers. Maier said APCO is working on interrogation criteria that ask for details "at such granular level that an AI cannot scoop that information from a public database." Even so, both officials emphasized that dispatchers will still treat a reported life-threatening emergency as real and send help. "It's important to understand that no matter what, public safety is going to have a very strong response," Heinze said, "even if there are indicators that it is a swatting incident."