Deepfakes: when AI puts words in anyone's mouth
A deepfake is an image, video or audio recording created or altered by AI to make it look as if a real person said or did something. Some are made for fun, but many are used to deceive, scam or harm.
Data as of October 22, 2026
Key figures
| Level | Beginner |
|---|---|
| Introduced by | Contraction de « deep learning » et « fake » (Office québécois de la langue française) |
A fake built with deep learning
The word deepfake combines “deep learning”, the technique behind most of today’s AI, and “fake”. The Office québécois de la langue française describes it as an audiovisual manipulation technique that uses deep learning algorithms to create ultra-realistic fakes (OQLF entry, in French).
The EU AI Act gives a legal definition: image, audio or video content generated or manipulated by AI that resembles existing people, objects, places or events and would falsely appear authentic. Written text is not covered by that definition.
What it looks like in real life
- The “family emergency” scam: a voice message mimics a relative and begs for an urgent bank transfer.
- The fake celebrity video promoting a “guaranteed” investment.
- At work, a fake call from an executive asks the finance team to pay an invoice right away.
- Harassment: faked intimate images shared to humiliate someone.
- Legitimate uses exist too: film dubbing, special effects, clearly labeled satire.
These schemes are covered in more detail in our entry on AI scams.
What EU law says
Article 50 of the EU AI Act requires anyone who uses AI to produce a deepfake to disclose that the content was artificially generated or manipulated (Article 50). For an evidently artistic, satirical or fictional work, the disclosure can be discreet so it does not spoil the work. The Act does not replace existing laws on defamation, fraud or image rights.
Good habits
Slow down. Urgency is a scammer’s main weapon.
Check through another channel. Call the person back on a number you know. Some families agree on a spoken safe word.
Don’t rely on detectors alone. AI content detectors and watermarks help, but none is fully reliable.
Don’t share a shocking video before checking where it came from.
To get comfortable with these ideas step by step, start with the Start with AI path.
Related terms
- AI scams · Frauds that use AI, such as cloned voices, fake videos or very convincing messages, to steal money or personal data.
- AI watermarking · A mark, often invisible, embedded in an image, audio, video or text to signal that it was generated by AI.
- EU AI Act · The European law that regulates AI systems according to their level of risk, with bans and transparency duties.
- Image generation · When an AI creates a brand-new picture from a written description (a prompt) or from another image.
- AI detector · A tool that estimates whether a text was written by AI; its verdict is a probability that can be wrong, never proof.
- Deep learning · Machine learning that uses neural networks with many layers; it is the technique behind today's chatbots, image and voice AI.
Full glossary
Frequently asked questions
Where does the word deepfake come from?
It blends “deep learning” and “fake”. The Quebec French language office (OQLF) notes that the term first described AI face-swapping, then broadened to cover all kinds of AI-made audiovisual fakes.
How can I tell if something is a deepfake?
There is no foolproof sign. Be wary of urgent requests, an unusual tone or an implausible setting. Verify through another channel: call the person back on their usual number or look for the original source of the video.
Are deepfakes illegal?
Not in themselves: a parody or a film can contain one. In the EU, however, the AI Act requires people who use AI to create a deepfake to disclose that the content was artificially generated or manipulated. Fraud, defamation and impersonation remain illegal, with or without AI, and rules vary by country.
What should I do if I'm targeted by a deepfake?
Keep evidence (screenshots, links, dates), report the content to the platform hosting it and, in case of fraud or serious harm, contact the police. For advice on your situation, talk to a lawyer or a victim support organization.
Data sources
Data as of October 22, 2026
Report an error
Does a figure, a date or a fact look wrong? Tell us: every report is checked.