Deepfakes, Consent and Likeness: The Lines You Should Never Cross

Deepfakes, Consent and Likeness: The Lines You Should Never Cross

User avatar placeholder
Written by Johnson

October 12, 2026

The same techniques that let you animate an original character can put a real person’s face and voice into a video they never made. That is the core of the deepfake problem. For creators using AI video, especially open tools without filters, understanding consent and likeness is not optional. It is the main thing standing between creative work and real harm to real people.

Deepfakes, Consent and Likeness: The Lines You Should Never Cross

What counts as a deepfake

A deepfake is synthetic or manipulated media that realistically shows a real, identifiable person saying or doing something they did not say or do. It can involve face swapping, full-body generation, lip sync to new audio, or voice cloning. The key elements are realism and a real person. Fictional characters, clearly stylised caricatures and your own likeness used by you are different situations.

Why it causes harm

  • Sexual deepfakes overwhelmingly target women and cause severe emotional, social and professional harm. They are a form of abuse.
  • Fraud: cloned voices and faces of executives or relatives are used to trick people into sending money.
  • Defamation and harassment: fake videos can damage reputations and be used to bully.
  • Misinformation: fake footage of politicians or events can mislead the public.
  • Erosion of trust: when anything can be faked, real evidence is doubted too.

The law is catching up

Many countries and states now have laws targeting non-consensual sexual deepfakes, impersonation for fraud, and deceptive election content. Some protect performers’ digital replicas and voices. Penalties can include criminal charges and civil damages. Platforms ban most harmful deepfakes and remove them on report. See Is Uncensored AI Video Legal?

Consent: what it really means

Real consent for using someone’s likeness or voice should be:

  • Explicit: a clear yes, not assumed from silence or friendship.
  • Informed: the person understands what will be made, how realistic it is and where it will appear.
  • Specific: consent for a training video is not consent for an ad or a different project.
  • Documented: for anything beyond private fun, written consent protects everyone.
  • Revocable where possible: agree what happens if they change their mind.
  • Given by an adult with the capacity to consent. Never create likeness content of minors beyond harmless, family-approved uses, and never anything sexual.

A public photo on social media is not consent. Fame is not consent. A person being dead does not automatically remove all rights either; some places protect deceased people’s likeness, and families can be harmed.

A simple consent agreement

For projects using a real person’s likeness or voice, a short written agreement should cover:

  1. who is giving consent and who receives it;
  2. the project and purpose;
  3. what will be created: face, voice, full body, level of realism;
  4. where it will be published and for how long;
  5. whether models such as LoRAs or voice clones will be kept, and who can use them;
  6. payment, if any;
  7. approval rights over the final content;
  8. how to withdraw consent and what happens then.

For commercial work, use a proper contract reviewed by a lawyer.

Legitimate uses with consent

  • an actor licensing their digital double for a film, with clear terms;
  • a company creating a presenter avatar of a willing employee for training videos;
  • a creator making an avatar of themselves for their own channel;
  • translating a speaker’s own video into other languages with their approval;
  • restoring and animating family photos for private memories with the family’s agreement.

Lines never to cross

  • Any sexual or intimate content of a real person without their explicit consent.
  • Any sexual content involving minors, real or fictional.
  • Putting false statements in real people’s mouths in a way that could deceive.
  • Impersonation for fraud, scams or harassment.
  • Training LoRAs or voice clones on people from their social media without permission.
  • “Undressing” or altering photos of real people.

Fewer filters do not change any of this. When you run models yourself, these lines are your responsibility to hold.

Satire and commentary

Satire of public figures has a long tradition and some legal protection in many countries. AI satire stays on safer ground when it is clearly exaggerated, obviously not real, labelled as parody and not sexual or defamatory. Realistic fakes presented without context can be shared without the joke and mislead people, which is why platforms restrict them heavily.

Protecting yourself from deepfakes

Creators are also targets. Limit high-quality voice samples and face references you post publicly, watermark original content, set up alerts for your name, and know the reporting processes on major platforms. If you are targeted by sexual deepfakes, many regions have support organisations and legal remedies; report to the platform and, where appropriate, to the police.

Designing projects around consent

The simplest way to avoid every likeness problem is to design projects around original characters. AI makes it easy to create memorable fictional people, and you own them creatively. When a real person’s presence genuinely adds value, involve them from the start, get proper consent and give them a say in the result.

If you receive a request that crosses the line

Freelancers and studios sometimes receive requests to create content of a real person without consent, often presented as a joke or a favour. Decline clearly and without debate. You can explain that you only work with consented likenesses and original characters. Keeping a short written policy on your website or rate card makes this easier and signals professionalism to legitimate clients.

Frequently asked questions

Is face-swapping my friend for a joke okay?

Only with their clear consent, and never in sexual, humiliating or deceptive ways. Many “jokes” cause real harm.

Can I use a celebrity lookalike?

If the result is intended to be mistaken for the celebrity, the same likeness issues apply. Courts in several countries have treated deliberate lookalikes in ads as using the celebrity’s identity.

Does labelling make a deepfake acceptable?

Labelling helps with deception, but it does not replace consent, especially for intimate content. A labelled non-consensual sexual deepfake still causes harm and is still illegal in many places.

Image placeholder

Want to suggest a topic, contribute a guide or discuss a partnership? Get in touch with our team.

Leave a Comment