What is a deepfake? Definition, examples and risks

A plain definition

A deepfake (also written "deep fake") is AI-made content that convincingly imitates a real person: their face, their body or their voice.

The EU AI Act gives the legal definition used across Europe: "AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful" (Article 3(60)).

What separates a deepfake from an old-fashioned photo edit is that software does the work after "learning" from pictures, videos or recordings of the person. No editing skills are needed: a few public photos or a few seconds of someone's voice can be enough.

Where the word comes from

The term surfaced in late 2017 on Reddit, where a user called "deepfakes" posted fake porn videos by swapping actresses' faces onto other people's bodies. Motherboard (Vice) reported on it on December 11, 2017. From day one, the word has been tied to non-consensual sexual content.

The main types of deepfakes

Real examples

• Video-call fraud. In early 2024, a finance employee at the engineering firm Arup in Hong Kong paid out about HK$200 million (around US$25 million) after a video conference in which the CFO and other colleagues were deepfakes.
• Sexual images at scale. The Center for Countering Digital Hate estimated in January 2026 that Grok's image tool on X produced around 3 million sexualized images in 11 days, including about 23,000 that appeared to depict children.
• Scam ads. The faces and voices of celebrities, doctors and ordinary people are used in fake investment and product ads. Your face in a scam ad?
• Voice cloning. McAfee researchers found that three seconds of audio was enough to produce a clone with an 85% voice match. What to do if your voice was cloned.

Why deepfakes are harmful

• For the people targeted. Most deepfakes online are sexual and almost all of those target women (98% and 99% in Security Hero's 2023 study). They are used for harassment, humiliation and sextortion.
• For money. Deepfakes are used to get past identity checks, to pose as a relative or a boss, and to sell fake investments.
• For trust. When anything can be faked, real footage can be dismissed as fake too. Researchers call this the "liar's dividend".
• Because they are hard to spot. In a 2025 iProov study of 2,000 people in the UK and US, only 0.1% correctly told apart all the real and fake content. How to spot a deepfake.
• Because they spread. A deepfake removed in one place can stay up on ten other sites and show up in Google under your name.

Are all deepfakes illegal?

No. The same technology is used in film, dubbing, satire and art. The problem is using a real person without consent, or to deceive.

• United States: the federal TAKE IT DOWN Act makes it a crime to publish non-consensual intimate images, including AI "digital forgeries", and since May 19, 2026 platforms must remove them within 48 hours of a valid request. Many states have their own deepfake laws.
• United Kingdom: sharing intimate images without consent, including deepfakes, is an offense, and since February 6, 2026 creating a sexual deepfake of an adult without consent is also an offense in England and Wales.
• European Union: since August 2, 2026, the AI Act requires anyone deploying a deepfake to disclose that it was AI-generated or manipulated (Article 50).

This is general information, not legal advice. Laws differ between countries and states: talk to a lawyer where you live. Deepfake laws explained.

If the person shown is under 18: use Take It Down (NCMEC) in the US or Report Remove (Childline and the IWF) in the UK, and report to the NCMEC CyberTipline. See our page for under-18s.