WEBVTT

00:00:18.307 --> 00:00:22.230
Imagine you're working at your job on a college campus.

00:00:22.307 --> 00:00:24.846
Unexpectedly, you get a phone call.

00:00:25.000 --> 00:00:27.653
It's from your child, supposedly.

00:00:27.730 --> 00:00:30.307
They're in distress, supposedly,

00:00:30.384 --> 00:00:34.461
and need you to send them money. So you do.

00:00:34.538 --> 00:00:36.038
But it's not your child.

00:00:36.038 --> 00:00:48.423
A bad actor used artificial intelligence to make a believable recorded copy of your child's voice from voicemail, perhaps a video, or social media.

00:00:48.500 --> 00:00:52.192
They then use that voice recording to try to trick you.

00:00:52.269 --> 00:00:59.576
This is called a voice cloning deepfake attack — a large and growing problem.

00:00:59.615 --> 00:01:10.307
From 2024 through 2033, the annual global growth rate for voice deepfakes is projected to be 37.6%.

00:01:10.384 --> 00:01:22.423
50% of educators say teachers or administrators in their school district have been misled by an AI-created deepfake, according to Education Week.

00:01:22.500 --> 00:01:26.538
Voice deepfakes are especially difficult to detect.

00:01:26.615 --> 00:01:37.884
Bad actors can easily disguise voice inconsistencies by using simultaneous background music or crowded restaurant noise to make detection harder.

00:01:38.038 --> 00:01:42.884
To better detect these crimes, you can take the following steps.

00:01:43.038 --> 00:01:46.884
First, listen for an odd tone of voice.

00:01:47.038 --> 00:01:51.153
Second, listen for interrupted voice sounds.

00:01:51.230 --> 00:01:57.576
Third, detect if the voice sounds robotic, choppy, or abrupt.

00:01:57.615 --> 00:02:06.230
Four, tune in for awkward pauses, words cut short, or unnatural speech patterns.

00:02:06.307 --> 00:02:10.653
Five, check on whether words are being mispronounced.

00:02:10.730 --> 00:02:18.230
Six, avoid posting long voice recordings on any publicly available sources.

00:02:18.307 --> 00:02:32.346
Seven, agree with your family members on a code word to ask each other that will confirm the authenticity of each other when calling. The bad actor won't know this code word.

00:02:32.423 --> 00:02:41.000
Take these smart steps before it's too late. Protect yourself from a barrage of voice deepfakes that could be coming your way.