← All resources

What is a deepfake?

A deepfake is synthetic media, video, image or audio, where AI puts a real person's face or voice into something they never did or said. For anyone who represents talent, that is the whole problem in one word: a convincing fake of a client, made without consent, spreading before anyone signed off on it.

How a deepfake is made

A model is trained on real footage, photos or voice of a person, then it either generates new media of them from scratch or swaps their face or voice onto existing footage. The first kind invents a scene that never happened. The second drops your client into someone else's video. Both used to need a specialist and a powerful computer. Now a phone app and a few minutes do a passable job, which is why the volume keeps climbing.

The kinds that actually hurt talent

Four do the real damage. Fake endorsements and scam ads use a trusted face to sell a fraud and take money from fans. Impersonation accounts pose as the person and speak for them. Non-consensual intimate deepfakes are the most harmful and the most urgent to remove. And unlicensed AI performances, a cloned voice on a track or in an ad, take income that should have been a paid deal.

Why it matters to whoever represents them

The law is catching up. The EU AI Act now requires deepfakes to be labelled as AI, and right-of-publicity law protects a person's likeness and voice. But a labelling rule does not find the fake, and a right you cannot prove was violated is hard to enforce. So the practical job is the same as ever: catch it, confirm it, and preserve proof that a specific fake existed at a specific time. That is what makes the law usable.

Common questions

Is every AI edit a deepfake?

No. Colour grading, cleanup and ordinary effects are not deepfakes. The line is impersonation: media built to look like a real person doing or saying something they did not.

Are deepfakes illegal?

It depends on the use and the country. Fraud, defamation, non-consensual intimate content and unlicensed commercial use of a likeness are already actionable in many places, and new AI-specific rules are arriving. The deepfake is a tool; the harm and the jurisdiction decide the claim.

Can a deepfake be detected?

Often, yes, with a mix of identity matching and AI-detection models plus a human check. No detector is perfect and the fakes keep improving, which is why detection is paired with fast, documented evidence.

Sources

Primary legislation and authoritative references. Laws change; confirm the current text and its application with qualified counsel.

  1. EU, AI Act (Regulation (EU) 2024/1689), Article 50 (deepfake transparency)
  2. US, CISA/NSA/FBI, Contextualizing Deepfake Threats to Organizations (2023)
  3. MIT Media Lab, Detect Fakes research project

General information, not legal advice. Likeness and image rights vary by country. Confirm specifics with qualified counsel.