Hacking. Disinformation. Surveillance. CYBER is Motherboard’s podcast and reporting on the dark underbelly of the internet.
“Hello, I just committed a crime and I want to confess,” a panicked sounding man said in a call to a police department in February. “I’ve placed explosives inside a local school,’ the man continued.
“You did what?!” the operator responded.
“I’ve placed explosives inside a local school,” the man said again, before specifying Hempstead High School in Dubuque, Iowa, and providing its address. In response to the threat, the school went on lockdown, and police searched the school but found nothing, according to a local media report.
The bombs weren’t real. But, crucially, neither was the man’s voice. The panicked man’s lines sound artificially generated, according to recordings of the swatting calls reviewed by Motherboard. It is unclear how exactly the caller generated the voice, be that some form of artificial intelligence tool or another speech synthesis program. The result, though, is a voice that sounds very consistent across multiple calls.
In fact, Motherboard has found, this synthesized call and another against Hempstead High School were just one small part of a months-long, nationwide campaign of dozens, and potentially hundreds, of threats made by one swatter in particular who has weaponized computer generated voices. Known as “Torswats” on the messaging app Telegram, the swatter has been calling in bomb and mass shooting threats against highschools and other locations across the country. Torswat’s connection to these wide ranging swatting incidents has not been previously reported. The further automation of swatting techniques threatens to make an already dangerous harassment technique more prevalent.
Swatting is when someone calls in a bogus threat in an attempt to direct law enforcement resources to a particular home, school, or other location. Often, swatting calls result in heavily armed police raiding an innocent victim’s home. At least one case has resulted in police killing the unsuspecting occupant.
Torswats carries out these threatening calls as part of a paid service they offer. For $75, Torswats says they will close down a school. For $50, Torswats says customers can buy “extreme swattings,” in which authorities will handcuff the victim and search the house. Torswats says they offer discounts to returning customers, and can negotiate prices for “famous people and targets such as Twitch streamers.” Torswats says on their Telegram channel that they take payment in cryptocurrency.
In the Hempstead High School case, authorities have charged a 16-year old boy who allegedly ordered the threats with making a threat of terrorism. But Torswats remains operational, publishing a steady stream of recordings of their crimes as recently as last week. Arguably, Torswats’ use of synthetic voices allows them to carry out swatting threats at scale with relatively little effort, while also protecting what their own voice sounds like.
Do you know anything else about how AI-powered voices are being abused? We’d love to hear from you. Using a non-work phone or computer, you can contact Joseph Cox securely on Signal on +44 20 8133 5190, Wickr on josephcox, or email [email protected].
Motherboard has determined that Torswats’ other targets include a CBD shop in Florida; the corporate headquarters of a Bethesda, Maryland, intelligence company that tracks extremism; and multiple private residences across Virginia, Massachusetts, Texas, California, and more.
“The FBI takes swatting very seriously because it puts innocent people at risk,” Steve Bernd, public affairs at FBI Seattle, told Motherboard in an email. Bernd said FBI Seattle was aware of the threats made against Hempstead High School. “These calls are dangerous to first responders and to the victims. The callers often tell tales of hostages about to be executed or bombs about to go off. The community is placed in danger as responders rush to the scene, taking them away from real emergencies, and the officers are placed in danger as unsuspecting residents may try to defend themselves.”
Motherboard’s reporting on Torswats comes as something of a nationwide swatting trend spreads across the United States. In October, NPR reported that 182 schools in 28 states received fake threat calls. Torswats’ use of a computer generated voice also comes as the rise of artificial intelligence poses even greater risks to those who may face harassment online. In February, Motherboard reported that someone had doxed and harassed a series of voice actors by having an artificial intelligence program read out their home addresses. Motherboard has also long reported on the threat posed by deepfakes, which are artificially generated videos of people, often without their consent. Deepfakes started as a tool to create non-consensual pornography of specific people.
On their Telegram channel, Torswats has uploaded at least 35 distinct recordings of calls they appear to have made. Torswats may have made many more swatting calls on others’ behalf, though: each filename includes a number, with the most recent going up to 170. Torswats also recently shuttered their channel before reappearing on Telegram in February.
In all of those 35 recordings except two, Torswats appears to have used a synthesized voice. The majority of the calls are made with a fake male sounding voice; several include a woman which also appears to be computer generated.
Torswats is seemingly able to change what the voice is saying in something close to real-time in order to respond to the operator’s questions. These sometimes include “where are you located,” “what happened,” and “what is your name?”
Motherboard listened to dozens of calls uploaded by Torswats and cross-referenced them with local media reports to determine some of the impact and geographical spread of the swatter’s threats. The media reports did not mention Torswats or the use of a computer generated voice, but Motherboard was able to link multiple reported swatting incidents to the recordings uploaded by Torswats. (Motherboard redacted some private information in some of the calls embedded in this piece.)
“In this specific case, the calls to 911 were placed with an automated or synthesized voice.”
In other cases, Motherboard used the address in the swatting call to identify the target. One example was SITE Intelligence Group, a private intelligence company that tracks extremism. In the Torswats uploaded call, the male synthetic voice claims they are making the threat in the name of the Order of the Nine Angles, a satanic neo-Nazi group. It is not clear if the claimed link to the group is genuine or misdirection. In March, SITE published a report about Telegram users who offer bomb threats and swatting calls for money. SITE did not respond to a request for comment.
In the case of Hempstead High School, Brendan Welsh, public information officer at the Dubuque Police Department, which responded to the incident, confirmed to Motherboard that an automated voice was used.
“In this specific case, the calls to 911 were placed with an automated or synthesized voice. On a grander scale though, this investigation continues as we believe our local student was working with another subject outside of our jurisdiction to plan and make these threats,” he told Motherboard in an email. Welsh added that the Cedar Rapids FBI field office assisted the investigation, and that FBI Seattle is working on a portion of the investigation as well.
Last month, VICE News reported that multiple schools in Pennsylvania were shut down after what were believed by authorities to be “computer-generated swatting calls.” It is not entirely clear if those threats were made by Torswats or another entity.
In August, Rep. Marjorie Taylor Greene claimed to have been swatted. A “computer generated voice” took credit for the call, according to a police report obtained by the Daily Dot.
At the end of March, authorities charged a 20 year old man, Ashton Connor Garcia, for allegedly making more than 20 swatting calls in the U.S. and Canada. The Associated Press reported prosecutors believe that Garcia used “voice-over-internet technology.”
Earlier this month, Torswats allegedly changed their tactics: they claimed to have made a swatting call using their own voice.
“One of the first swattings I’ve done with my own voice in a long time ???,” they wrote on their Telegram channel. Motherboard cannot verify if this audio does indeed contain Torswats’ own voice, but the call does not appear to be using a synthesized voice.
In the subsequent recording, they start with much the same script as their automated voice. “I’ve done something really bad and want to kill myself,” they tell the operator. They then claim they came out to their parents as a transgender woman, that they have an AR-15, and will shoot any police who respond.
“Forgot to cut off my laugh at the end,” Torswats wrote on Telegram.
After publication of this article, Torswats deleted the audio recordings from their Telegram channel and claimed they were stopping the service for at least one month. “Time to dip a bit,” they wrote on the channel.
Update: This piece has been updated to include that Torswats has now deleted the recordings.