AI hoaxes are deceiving 911 dispatchers, and they're only getting harder to detect
In August, a caller made a bomb threat to Wheatland Union High School in Yuba County, Calif.
Police responded overnight and classes were canceled.
asserted
classes → deceive → County
There was no bomb and no one was hurt.
Authorities said it was part of a series of hoax calls — also called swatting — made to schools across the country over the summer.
asserted
it → be → summer
Swatting calls can be made to 911 or police departments, with callers reporting a bomb threat, mass shooting or some other kind of life-threatening emergency with the intention of initiating a massive emergency response, according to the FBI.
911 dispatchers have long dealt with this problem.
uncertain
dispatchers → swat → problem
But the incident in Yuba County was different: The call was made by an AI-generated voice, a new and growing challenge for emergency responders.
asserted
call → make → responders
"Those voices [of swatting callers] used to be pretty easy to recognize, but now they're natural sounding.
asserted
they → swat → callers
They can insert pauses.
asserted
They → insert → pauses
They can use words like, 'uh', or 'maybe, um, let me think about it.'
asserted
me → use → it
And it sounds very natural," says Mel Maier, the CEO and executive director of APCO International, the Association of Public Safety Communication Officials.
asserted
Maier → sound → Officials
The goal of swatting calls is to harass or retaliate against the targets of the calls — like a public official.
asserted
goal → swat → official
They usually trigger a big, chaotic response from police, redirecting public safety resources from real emergencies.
asserted
They → trigger → emergencies
People also have been killed and injured in swatting attacks, Maier notes.
asserted
Maier → kill → attacks
Swatting calls made by people using AI-generated voices or background sounds are making these hoax calls sound even more realistic, and harder to determine what's real and what isn't.
asserted
what → swat → voices
The 9-1-1 Association, which both represent 911 dispatchers, say incidents like the one in Yuba County reflect their growing worries over how AI could be used by bad actors.
uncertain
AI → represent → actors
APCO and the Public Safety Threat Alliance, an information-sharing and cybersecurity group, looked into the prevalence of swatting to 911 call centers.
asserted
APCO → look → centers
They published their findings in June based on 766 responses from public safety communications professionals.
asserted
They → publish → professionals
About 14% of dispatchers said they were targeted using AI tools.
asserted
they → say → tools
Maier says he thinks that number is going to keep going up with increasingly advanced tools
Signs of a swatting call
Companies like ElevenLabs, an artificial intelligence company creating synthetic voice software, offer AI tools that are incredibly cheap and readily available, according to Brandon May, an assistant professor of forensic psychology at the Florida Institute of Technology.
uncertain
that → say → Technology
"I'm able to clone my voice on ElevenLabs for less than $20, and I can change the intonations.
asserted
I → clone → intonations
I can change how deep or high-pitch my voice is.
asserted
voice → change → ?
I can change the pauses," he says.
asserted
he → change → pauses
And this makes it hard for anyone to detect these AI fakes.
asserted
anyone → make → fakes
"It is no longer simple to just look at an image or hear a video or hear a voice and go, 'That is AI.'
asserted
That → look → voice
People are duped by this all the time," he says.
asserted
he → dupe → this
Elaborate fraud callers using AI chatbots are even tricking people into handing over their life savings, May says.
uncertain
May → use → savings
And the increasing sophistication of these fake voices and sound effects makes it especially difficult for emergency dispatchers to deduce what is real or not during what May calls "high-stakes, high-consequence" calls.
uncertain
May → increase → what
"It is going to be nearly impossible to detect," he says.
asserted
he → go → ?
Despite the difficulty, APCO and NENA are developing ways to train their members to deal with these new threats.
asserted
APCO → develop → threats
April Heinze, NENA's vice president and chief of 911 operations, says there are some hallmarks of a swatting call.
asserted
Heinze → say → call
For example, most swatters do not dial 911 directly.
asserted
swatters → dial → 911
"The reason being is 911 provides location.
asserted
911 → provide → location
It provides a callback number.
asserted
It → provide → number
It provides the name of the person who owns the phone.
asserted
who → provide → phone
It provides a lot of extra data, and it's really impossible to spoof your location on 911," she says.
asserted
she → provide → 911
Swatting calls are usually made to police departments or courthouses, Heinze explains.
asserted
Heinze → swat → departments
These callers are then ultimately transferred into the 911 centers.
asserted
callers → transfer → centers
So trying to trace those calls is impossible, she says.
asserted
she → try → calls
Training evolves to address new AI threat
Both Heinze and Maier decline to describe the specifics of training to identify potential AI swatting calls.
asserted
Heinze → evolve → calls
The more information available, they say, the easier it will be to train AI to fool dispatchers going forward.
asserted
it → say → dispatchers
Maier does note one idea for training.
asserted
Maier → note → training
APCO is working to create training criteria on caller interrogations where the dispatcher asks the caller specific questions at "such granular level that an AI cannot scoop that information from a public database, cannot capture it from a social media post, or simply make guesses," he says.
asserted
he → work → guesses
…and 3 more, not listed.
The Aporia
