1. What is a deepfake?

A deepfake is digitally generated or manipulated content that can make a person appear to say or do something they never said or did. It may involve a face, voice, photograph, video or other identifying features.

AI voice cloning can imitate a person's speech. Image-generation systems can produce realistic-looking photographs of fictional events. Video tools can combine a real person's appearance with generated speech or movement.

Not every digital replica is harmful. Filmmakers may use authorised replicas for special effects, performers may license their likeness, and people may use voice technology to support communication. The key questions are consent, context, transparency and the potential impact on the person represented.

A convincing recording should not automatically be treated as authentic. Equally, genuine material should not automatically be dismissed as fake just because AI can now imitate people.

Source: Police.uk — Deepfakes.

2. Why digital identity has become a public issue

For many years, creating a convincing imitation of someone's voice or appearance required specialist skills, time and money. Generative AI has lowered those barriers, allowing more people to create realistic replicas using widely available tools.

This creates opportunities for creative businesses, but it also makes impersonation easier to scale. A fake recording can be shared across multiple platforms, copied by other users and detached from its original context.

The consequences can include financial loss, distress, bullying, damage to professional relationships and the spread of false information. Ordinary people can be targeted just as public figures can.

The Government's March 2026 report on copyright and AI recognised that digital replicas can cause harm even where existing legal protections do not provide a straightforward solution. It proposed further exploration of options to clarify legitimate uses and address unacceptable imitation.

Source: UK Government — Report on Copyright and Artificial Intelligence.

3. AI voice cloning and financial fraud

A cloned voice may be used to imitate a family member, colleague, manager, artist or business representative. Someone might receive an unexpected call or voice message claiming that urgent money is needed, a bank account must be changed, or confidential information should be provided.

Voice recognition alone is not a reliable way to verify a person's identity when AI imitation may be involved. A familiar voice can sound convincing without proving who is speaking.

Individuals and organisations can reduce risk by agreeing a separate verification procedure. For example, a family or workplace can use a known telephone number, a second communication channel or a pre-agreed check before transferring money or sharing sensitive information.

Businesses should ensure that payment instructions and changes to bank details are checked through an established process, even when a request appears to come from a senior colleague.

The principle is simple: a voice recording is evidence to consider, not proof of identity by itself.

4. When a fake image becomes abuse

One of the most serious uses of generative AI is the creation of intimate or sexual images of people without their consent. A real photograph can be manipulated, or an entirely synthetic image can be generated to appear to depict a real person.

Such material can be used to humiliate, threaten, intimidate or control someone. The fact that an image is fabricated does not make the harm imaginary.

UK law has developed specific protections. Since February 2026, offences under the Sexual Offences Act 2003, as amended by the Data (Use and Access) Act 2025, cover creating or requesting the creation of certain purported intimate images of adults without consent or a reasonable belief in consent.

The Crime and Policing Act 2026 includes further measures targeting the making or supply of tools designed to generate purported intimate images without consent. It also provides for new platform duties concerning the removal and prevention of recirculation of qualifying non-consensual intimate images.

These provisions have specific legal definitions and conditions. Not every AI-generated image falls within the same offence, so people affected should seek appropriate advice rather than assuming that a lack of realism or the use of AI removes legal protection.

Sources: Crown Prosecution Service — Rape and Sexual Offences Prosecution Guidance and Crime and Policing Act 2026: Intimate Image Abuse Factsheet.

5. Deepfakes, children and safeguarding

Children and young people may be especially vulnerable to impersonation, online bullying, coercion and the circulation of fabricated images. A fake image can be used to shame a pupil, damage friendships or pressure someone into doing something they do not want to do.

Adults working with children and young people should take reports seriously, avoid blaming the person depicted, and follow their organisation's safeguarding and reporting procedures. Staff should not ask a child to circulate harmful material as proof or forward intimate images unnecessarily.

Where an image may be illegal, staff should follow safeguarding guidance and seek advice from the designated safeguarding lead or appropriate authorities about preserving evidence safely. Do not download, duplicate or redistribute suspected child sexual abuse material.

Schools, youth organisations and creative projects should have clear procedures for image consent, photography, online publication, incident reporting and escalation. Young people should know whom they can approach if an image of them is manipulated or shared without permission.

Source: Home Office and Ministry of Justice — Intimate Image Abuse Factsheet.

6. What does the Online Safety Act do — and what are its limits?

The Online Safety Act 2023 establishes duties for certain online services to address illegal content and protect children from specified harmful content. AI-generated content is not automatically outside the rules simply because it was produced by a machine.

However, the precise scope depends on the service, the content and the legal duties that apply. Ofcom explained in February 2026 that some standalone AI chatbots did not fall within the Online Safety Act's existing scope in the same way as user-to-user platforms or search services. That created limits on what the regulator could investigate under the law at that time.

Ofcom's 2026 investigation into X followed concerns about sexual deepfakes generated and shared using Grok. Ofcom examined whether X had met its duties to assess and mitigate risks and remove relevant content. The investigation also highlighted the difference between regulating a platform where content is shared and regulating a standalone system that generates it.

This distinction matters. Protecting people may require attention not only to the person who creates a harmful image, but also to the systems that generate it, the services that distribute it and the processes available to remove it.

Source: Ofcom — Investigation into X and the scope of the Online Safety Act.

7. Who controls a person's voice and likeness?

A person's face and voice can be commercially valuable. Actors, musicians, presenters, voice artists and other performers may license their likeness or voice for specific projects. AI can make such arrangements more flexible, but it can also create replicas outside the terms of an agreement.

The UK does not currently have one general personal image right that automatically gives every individual control over every digital depiction of themselves. Different legal protections may apply depending on the circumstances, including data protection, intellectual property, defamation, contractual rights and criminal law.

Copyright does not automatically protect every aspect of a person's identity. A person may be imitated without the replica copying a substantial part of a particular copyrighted recording or film. This is one reason why digital-replica protection has become part of the wider policy debate.

In July 2026, the Government told Parliament that it recognised the risks of malicious AI-generated voice and likeness cloning and was considering whether further measures were needed. Its March 2026 copyright report also discussed digital replicas and possible approaches to protecting people from unacceptable imitation.

Sources: UK Parliament — Written answer on AI voice and likeness cloning and Report on Copyright and Artificial Intelligence.

8. Consent must be specific, informed and meaningful

Consent to one use of a person's image or voice should not automatically be treated as permission for every future use.

For example, agreeing to have a song recorded is not necessarily the same as agreeing to have one's voice cloned for new dialogue, advertisements or synthetic performances. Agreeing to appear in a film is not automatically agreement to create a digital replica for unrelated projects.

Creative contracts should clearly describe the intended use, the material involved, the duration and scope of the permission, any payment, and whether further uses require additional agreement.

People should have a genuine opportunity to understand what they are agreeing to. Consent procedures should be accessible, particularly for disabled and neurodivergent participants who may need information in alternative formats, additional processing time or support to make an informed choice.

For projects involving children or adults at risk, consent must be considered alongside safeguarding responsibilities, organisational policies and the person's welfare. A signed form alone does not remove the need for ongoing care and appropriate oversight.

9. How can people recognise and respond to a deepfake?

There is no single visual or audio clue that reliably identifies every deepfake. Some manipulated media contains obvious inconsistencies, while other examples may be difficult to distinguish from genuine recordings.

Instead of relying on appearance alone, people can check the wider context:

• Look for the original source rather than relying on a repost or screenshot.
• Check whether reputable independent sources confirm a significant claim.
• Contact the person or organisation through a known, separate channel if identity matters.
• Be cautious about urgent requests for money, passwords, security codes or confidential information.
• Check dates, context and whether an older clip has been presented as a new event.
• Preserve relevant URLs, account details, dates and messages if you need to report impersonation.
• Do not redistribute suspected harmful material merely to warn other people.

AI-detection tools can sometimes provide useful clues, but their results are not definitive proof. A responsible assessment considers provenance, context, independent evidence and the limitations of the detection method.

10. What should you do if someone creates a harmful replica of you?

If someone has impersonated you or shared a harmful AI-generated image, voice recording or video, consider the following steps:

1. Prioritise safety. If there is an immediate threat, contact emergency services. If a child or vulnerable person is involved, follow appropriate safeguarding procedures.
2. Record the evidence safely. Keep URLs, usernames, timestamps, messages and screenshots where lawful and safe to do so. Do not download or redistribute illegal imagery.
3. Report the content. Use the platform's reporting process and select the most relevant category, such as impersonation, harassment or non-consensual intimate imagery.
4. Contact the relevant authorities where appropriate. Depending on the circumstances, this may include the police, a safeguarding lead, a regulator or a legal adviser.
5. Protect your accounts. Change compromised passwords, enable multi-factor authentication and alert relevant contacts if your identity is being used for fraud.
6. Seek support. Harmful impersonation can be distressing. You do not have to manage the situation alone.

In the UK, suspected crime can be reported to the police. For online content, Ofcom provides information about its regulatory role and the kinds of complaints it can consider. The appropriate route depends on the issue and the service involved.

Useful links: Police.uk, Ofcom Online Safety and Intimate Image Abuse Factsheet.

11. What should creative businesses and organisations do?

Media companies, music labels, schools, charities and community organisations should establish clear rules before using AI-generated voices, faces or other digital replicas.

• Obtain specific, documented permission for intended uses of a person's voice, image or likeness.
• Explain whether AI will generate, alter, extend or reproduce a person's identity.
• Separate permission for an initial project from permission for future or unrelated uses.
• Keep agreements, releases, source files and records of approved uses secure.
• Check the terms of AI tools and the rights attached to any uploaded material.
• Use clear labelling or disclosures where required or appropriate to prevent audiences being misled.
• Set up a straightforward process for reporting concerns, correcting mistakes and requesting review.
• Train staff and contractors to recognise impersonation, fraud and non-consensual content.
• Include safeguarding and accessibility in decisions involving children, disabled people and adults at risk.

Responsible practice is not just about avoiding legal problems. It is about preserving trust between creators, performers, participants, audiences and the organisations using the technology.

12. Following the money: who benefits from digital replicas?

Digital replicas sit within a growing commercial ecosystem. AI developers may sell generation tools, platforms may attract engagement, and entertainment companies may use licensed digital performers or voices to produce new content.

Some performers may benefit from carefully negotiated licensing agreements. Others may face unauthorised imitation, loss of control over their identity or the expense of trying to remove harmful content.

To understand the economic impact, policymakers and the public should ask:

• Who receives payment when a digital replica is commercially used?
• What evidence demonstrates that the person gave informed permission?
• Can a person understand, challenge or end an agreement where appropriate?
• Who is responsible when a tool is used to facilitate fraud or abuse?
• How quickly can harmful material be reported and removed?
• Can independent creators access the same practical protections as large companies?
• Are the rules accessible to people with different communication and support needs?

The goal should be an environment in which legitimate creative uses can develop while people retain meaningful protection against harmful impersonation.

Conclusion: your identity should not become someone else's product without your knowledge

AI-generated voices, faces and videos can support creativity, accessibility and entertainment. They can also be used to deceive, exploit and harm people.

The UK has strengthened its response to certain forms of intimate-image abuse and continues to examine the wider challenges posed by digital replicas. But legal protections vary according to the circumstances, and the development of new tools continues to test existing systems.

Consent, transparency, safeguarding and accountability should be part of the design and use of AI, not added only after something goes wrong.

For creators and organisations, the practical starting point is to make permissions clear, protect personal information, document decisions and provide a route for concerns to be addressed.

Technology may be able to reproduce a face or voice. That does not mean it should be free to misuse a person's identity.

Sources and further reading

• Report on Copyright and Artificial Intelligence — GOV.UK, March 2026
• Crime and Policing Act 2026: Intimate Image Abuse Factsheet — GOV.UK
• Rape and Sexual Offences Prosecution Guidance — Crown Prosecution Service
• Investigation into X and the scope of the Online Safety Act — Ofcom
• Parliamentary answer on AI voice and likeness cloning — 14 July 2026
• Deepfakes — Police.uk
• Crime and Policing Act 2026 — legislation.gov.uk

Research checked: 29 October 2026. This article provides general information, not legal advice. The scope and commencement of individual legal provisions can vary; consult current legislation and specialist advice for a specific case.