Safety Instinct

Navigation


Articles · Deepfakes · 2 min read

What is a deepfake?

A deepfake is AI-made video, images, or audio that convincingly imitates a real person — saying or doing things they never did. Here's how it works, and how to catch it:

Illustration of a cybercriminal creating a deepfake
1

What it is

AI trained on someone real

Feed software enough photos, clips, or voice samples of a person and it can generate brand-new footage of them — saying and doing things that never happened. The result can look and sound convincing.

2

The fake video

A familiar face, out of nowhere

A clip that looks real at a glance: a celebrity endorsing an investment, a "leaked" recording of someone in the news, or a video call from a colleague who isn't really there. The face moves and talks like the real person.

3

The cloned voice

A few seconds is enough

A short snippet of audio is all it takes to copy someone's voice. It turns up in panicked phone calls — "it's me, I'm in trouble, send money now" — using a voice you'd swear you recognise.

4

The tells

Check the source, not the clip

Look for unnatural edges around the face, odd blinking, or lips that don't quite match the words. But the real check is the source: does a channel you trust confirm it? Verify the person, not the recording. Seeing and hearing are no longer proof — confirm anything urgent through a channel you trust.

The technology will only get better at fooling your eyes and ears. When something urgent arrives, verify the person — not the recording.

Also worth knowing

Illustration of a scammer

Scammers · 4 min

How to identify a scammer