Deepfake
Also called Deepfake
A convincingly fake image, video, or voice generated by AI to show someone doing or saying something they never did.
Think of it like
A flawless forged photograph — except now the forger works in minutes and the seams are nearly invisible.
Example
A cloned voice of a company’s CEO calls the finance team and authorizes an urgent wire transfer that never should have happened.
How it actually works
Generative models can now synthesize faces, voices, and video realistic enough to fool people at a glance. Deepfakes power fraud, non-consensual imagery, and political disinformation, and detection is a losing arms race — as detectors improve, so do the fakes. The more durable defenses lean on provenance (watermarking, signed authentic media) and on plain skepticism about unverified footage, rather than on spotting the fake after the fact.
For product teams
A trust-eroding threat that raises the bar on verifying anyone’s identity or any piece of media.
For engineers
Realistic synthetic audio/video; post-hoc detection is an arms race, so provenance (watermarks, signed capture) is the more durable defense.
Related
- Watermarking — A provenance defense against it.
- Disinformation — Deepfakes supercharge it.
- Responsible AI — An ethics concern for deployment.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome