Deepfake Scams Hit 1 in 15 Australians: Familiar Voices Lie
Quick answer
Can you tell if a voice or video is a deepfake?
Often not. A review of 19 studies covering 24,000 people found about one in four could not spot a deepfake, and some voice-detection software has scored as low as 38.2% accuracy. Visual oddities may help a little, but the safer step is to hang up and call back on a saved number.
Picture a call from your daughter. It is her voice, her little laugh, her usual way of saying "Mum." She needs money, fast. Here is the problem: a deepfake (synthetic media created by AI, such as a fake voice, photo or video made to look and sound like a real person) can now fake all of it.
A deepfake can copy a face or voice you trust, so treat any urgent call or message as a question to check, never as proof.
Deepfake harm in Australia: what the numbers say about victims
Let's start with your safety, since that is why you are here. You are not being silly to worry. Research out of RMIT University, carried by Yahoo News Australia, found that about one in 15 Australian adults (7.7%) has been targeted by AI-made image abuse since turning 18. That is not a rare event. That is probably someone in your street, your office or your family group chat.
That number matters more than the first one. This is not a harmless prank. Most people hit by it say it changed their daily life. And the researchers note that most victims of sexually explicit fakes are women, so this is often targeted harassment, not random bad luck. Deepfakes are also a privacy violation, because they use a person's face without consent.
Why victims in Australia are not just "people who got fooled"
It is tempting to think victims were careless. Look at the detection numbers instead. A review of 19 studies covering 24,000 people, written up in The Conversation, found about one in four people cannot spot a deepfake. It also found people wrongly call real content a deepfake. So we are bad at both jobs: catching fakes and trusting real things.
Can you tell if a deepfake is fake? Why a familiar voice is not proof
Short answer: often, no. Deepfake detection software has scored as low as 38.2% accuracy for voices, according to figures gathered by Spiralytics. Some tools struggle, and your ear at 11pm, half-asleep, with a scared "relative" on the line, may struggle too. This article is part of a series, start with Facial Recognition False Arrest 9 Wrong Alerts 1 38m Faces.
You may have heard the old advice: look for odd signs, like skin that appears too smooth, blurry teeth or strange blinking. Sure, check. But treat those clues as a bonus, not a safety net. Deepfake videos keep improving, and a voice call has no skin to inspect at all. Deepfakes are also easy to pass around on social media, so a fake video can reach thousands of people before anyone checks it.
A familiar voice should prompt skepticism, not trust.
Expert research summary prepared for this story, drawing on TechTimes reporting on voice cloning
Deepfake impersonation: when a fake face sells you a fake investment
Deepfake impersonation (pretending to be a real person using AI) is already costing real money. The Australian regulator ASIC tied $7.4 million in losses to deepfake investment scams that used the Prime Minister's face. And this week brought charges against three men in an alleged $129 million AI voice-cloning fraud scheme. Deepfake scams run on the same trick, at bigger scale.
Why does this feel so close to home? Psychologists call it the availability heuristic: the stories we remember easily feel more likely. Right now the stories are everywhere, so the fear is not irrational. It is information.
Deepfake detection vs. real life: what a fake shows and what a real person gives you
| What you rely on | The problem now |
|---|---|
| A familiar voice | A deepfake can clone it; deepfake detection software has scored as low as 38.2% |
| A face on a video call | A deepfake video can look real; about one in four people cannot spot these fakes |
| Your gut feeling | People also call real video a deepfake, which makes trust shaky both ways |
| A call-back to a number you already saved | Much harder to fake, so this is the check that holds up |
Why awareness matters right now
- Money: cloned voices and faces power scams worth millions.
- Dignity and privacy: fake explicit images harm most victims in daily life.
- Evidence: if a photo or video can be faked, a perfect match may mean little without knowing where it came from.
What can you actually do about a deepfake call or message?
Do one thing: break the moment. Hang up. Call the person back on the number you already have saved, not the one that called you. Better still, agree on a family code word now, over dinner, before you need it. A scammer can copy a voice, but not a secret they never heard. Previously in this series: Facial Recognition Wa Police Scan 1 4m Faces Arrest 122.
If you have ever wondered whether a photo or profile is really who it claims to be, that is the exact question face comparison tools exist to answer. It is the same idea at work: do not ask "does this look right?" Ask "where did this come from, and can I check it?" The week's police facial recognition debate says the same thing from the other side. A computer's face match is a lead to question, not a verdict.
A deepfake makes a familiar voice or face a claim you have to check, and a call-back on a saved number is the simplest way to check it.
Recap: Fake voices and faces can now copy people you trust. Treat urgent requests as questions, and verify them another way.
So here is the question to sit with tonight: if the voice on the phone is perfect, what is the one thing only the real person would know?
Deepfake: Frequently Asked Questions
How many Australia deepfake victims are there?
The RMIT study reported by Yahoo News Australia found 7.7% of Australian adults, about one in 15, have faced AI-generated image abuse since turning 18. Of those victim-survivors, around 87% said it affected their health, work, relationships or finances. The figures cover image-based abuse specifically, so the full count of people hit by cloned voices or investment scams is likely higher, but this research does not measure that. Up next: Deepfake 1 In 15 Australians Hit And Familiar Voices Lie.
Can people tell if a video is real or fake?
Not reliably. A review of 19 studies covering 24,000 people, reported by The Conversation, found about one in four cannot identify a deepfake. People also mislabel genuine video footage as a deepfake. That is why checking where something came from beats trusting your eyes. Spotting odd details in deepfake videos helps a little, but it should never be your only defence against a convincing fake video.
Does skin appear too smooth in a fake video?
Sometimes. Older advice says to watch for skin that appears too smooth, odd blinking or blurry edges in deepfake content. These clues can help, but newer fakes often hide them. A cloned voice on a phone has no picture to inspect at all. Use visual clues as a bonus and rely on a call-back or a family code word for real protection.
What is deepfake impersonation?
Deepfake technology uses artificial intelligence to copy a real person's face or voice, and deepfake impersonation is using that copy so you believe you are dealing with them. Examples in this week's coverage include deepfake scams using the Prime Minister's likeness, which ASIC tied to $7.4 million in losses, and an alleged voice-cloning fraud scheme worth $129 million. The safest response is always to verify through a second, trusted channel.
How accurate is voice deepfake detection software?
It varies. Figures gathered by Spiralytics show voice detection accuracy as low as 38.2%. That shows some software can struggle to detect deepfakes that use cloned voices, including deepfake audio. Tools built on machine-learning algorithms only learn from fakes they have already seen, so newer fakes can slip past them. Detection will likely improve, but for now a human habit, like hanging up and calling back on a saved number, is a useful additional check.
Ready for forensic-grade facial comparison?
Full forensic reports with detailed similarity scoring. Results in seconds.
Run My First SearchMore News
AI Voice Cloning: Fake Call Takes Rs 1.2 Lakh From Florist
A Kolkata shop owner heard a voice she'd known for years and sent money fast. It wasn't that person. Here's the simple pause that beats AI voice cloning.
privacyTikTok's Response: Nearly 50 Nude Deepfake Apps, 'Not Ads'
A report says nearly 50 apps that make sexual deepfakes were pushed through TikTok ads. TikTok says the examples weren't ads. Either way, a familiar feed is not a safety guarantee.
facial-recognitionFacial Recognition in Stores: False Match Ejects UK Shopper
A shopper was thrown out of a UK store after a face-match alert said she'd stolen meat. She says she hadn't been there in months. Here's why an alert isn't proof, and what to say if it happens to you.