AI Voices Complicate 911 Bomb Threat Verification
Dispatchers report synthetic audio in swatting calls is rising, prompting new training protocols to distinguish real emergencies from digital hoaxes.
By Fatou Girard ·

A bomb threat against Wheatland Union High School in Yuba County, California, arrived in August via an AI-generated voice. Police cleared the building overnight, and classes were canceled for the day. No bomb was found, and no students were hurt. Authorities identified the call as part of a summer series of hoax calls targeting schools nationwide. The Yuba County incident marked a shift in how these calls are delivered, introducing synthetic audio that sounds indistinguishable from a human speaker.
Swatting involves calling 911 or a police department to report a false emergency, such as a bomb threat or active shooter. The goal is to trigger a massive response that diverts resources and harasses the target. While dispatchers have handled these calls for years, the integration of artificial intelligence presents a new technical challenge. Mel Maier, CEO and executive director of APCO International, the Association of Public Safety Communication Officials, noted that older AI voices were easy to spot. Modern versions insert natural pauses and filler words like 'um' and 'uh,' making them difficult to detect by ear.
The prevalence of this issue is documented in a June report by APCO and the Public Safety Threat Alliance. The survey included 766 responses from public safety communications professionals. Approximately 14 percent of dispatchers reported being targeted by calls using AI tools. Maier stated that he expects this percentage to rise as the technology becomes more accessible and sophisticated.
Brandon May, an assistant professor of forensic psychology at the Florida Institute of Technology, explained that the barrier to entry for creating these calls is low. He cited ElevenLabs, a software provider, noting that users can clone a voice for less than $20. May said he can adjust intonation, pitch, and pauses easily. This affordability means anyone can generate a convincing fake voice, complicating the detection process for emergency workers.
April Heinze, vice president and chief of 911 operations for NENA, the 9-1-1 Association, pointed to specific patterns in how these calls are made. Most swatters do not dial 911 directly because the system provides location data and callback numbers that are difficult to spoof. Instead, callers often dial police departments or courthouses. These calls are then transferred to 911 centers, making it nearly impossible to trace the original source. Heinze said this routing strategy is a hallmark of a swatting attempt.
In response to the growing threat, APCO and NENA are developing training methods for dispatchers. Both organizations declined to share specific technical details, citing the risk that detailed information could help bad actors refine their AI tools. Maier described one training approach involving granular caller interrogations. Dispatchers are being trained to ask questions that require specific, non-public knowledge. The aim is to pose questions that an AI cannot answer by searching public databases or social media posts.
Despite the rising difficulty, both Maier and Heinze emphasized that dispatchers will continue to treat every life-threatening report as real. Heinze stated that public safety must maintain a strong response even when indicators suggest a hoax. She explained that since callers report a life-or-death situation, emergency services cannot afford to take chances with their decisions.
The combination of cheap software and advanced voice synthesis is changing the daily work of emergency responders. Dispatchers are now tasked with identifying subtle digital artifacts in high-stress situations. The training programs are ongoing, but the tools used by callers continue to evolve at a pace that keeps the verification process complex.






