Esther Gunaratnam
- Partner
- IP & Commercial
The deepfake challenge: protecting creativity, identity and reputation
As AI-generated content becomes increasingly convincing, deepfakes are posing growing risks for artists, performers, collectors, celebrities and other high-profile individuals whose identity and reputation form part of their commercial value.
AI is accelerating the creation of synthetic media, including deepfakes, which replicate voices, images and performances with increasing realism. This creates commercial and legal risks for creators and the businesses which work with them. The English legal framework does not map neatly onto deepfakes which can cause challenges in identifying and pursuing a legal remedy.
This tension is recognised at policy level. The UK Government’s March 2026 report on copyright and AI considers the need to protect creative industries, and pauses reform pending further evidence.
A broad strategic approach which blends multiple forms of intellectual property and contractual protection will help manage risks.
WHAT ARE DEEPFAKES AND WHY DO THEY MATTER?
Deepfakes are synthetic media which falsely appear authentic. There are two key concerns for creative talents:
- their works may be used in AI training datasets without permission; and
- their identity, voice, likeness or elements of their creative output may be replicated or imitated in AI-generated outputs.
This may lead to unauthorised use of content, misattribution, distortion of works and false endorsement, as well as having a detrimental impact on their image and the value of their genuine work.
KEY RISKS AND OPPORTUNITIES
False endorsement can create reputational harm and brand dilution. Creators and brands may lose control over how their identity is commercially exploited. Existing sponsors, collaborators or partners may be concerned that the value of existing relationships is diminished. Remedial PR measures can be costly and time-consuming.
A business which uses an AI product could face uncertainty because the training datasets or AI-generated outputs may be unlawful.
A broader challenge is that enforcement of legal remedies is difficult due to lack of transparency over who created the deepfakes, evidential burdens and cross-border issues (many AI developers and deepfake generators are based outside the UK).
On the flip side, AI creates new markets for monetising creative works and aspects such as voice and likeness. Authorised licensing supported by contracts can provide individuals with control and additional streams of revenue.
COPYRIGHT AND AI TRAINING
Under the Copyright, Designs and Patents Act 1988, copyright owners have the exclusive right to copy their work; ‘copying’ includes electronic storage. AI training may infringe this right where a substantial part of a protected work is copied without permission. This is fact-specific.
There is an exception which allows text and data mining for non-commercial research; this requires lawful access. This exception does not permit general commercial AI training.
The High Court’s judgment in Getty Images v Stability AI illustrates the practical difficulties of bringing AI-related IP infringement claims. Getty alleged that Stability AI had used Getty’s copyright-protected images to train its AI model without permission and sued Stability AI in the English Courts. However, as copyright is territorial and Getty ultimately accepted that there was insufficient evidence that the relevant training activities had taken place in the UK, it abandoned its primary copyright infringement and database rights claim regarding AI training. That left the secondary infringement claim relating to the AI model weights. The English Court rejected this claim because the AI model weights did not store or reproduce (i.e. did not “copy”) Getty’s images. The case therefore highlights the evidential and jurisdictional challenges that rights holders may face when seeking to establish where AI training took place and which legal regime applies. It also left unresolved the broader question of whether, and in what circumstances, the use of copyright works for AI training infringes copyright under English law.
AI-GENERATED OUTPUTS
AI outputs may themselves infringe copyright if they reproduce a substantial part of a protected work. As with copyright infringement generally, it must be shown that the protected work was in fact copied rather than merely created independently. This is an area that can be difficult to substantiate in the context of AI output,
Academic analysis similarly frames liability around whether the output reproduces elements of the author’s intellectual creation.
It is important to understand exactly how AI is used to generate outputs. For example, if a brand engages an advertising agency to prepare content, the brand should consider how the agency uses AI, the protections it has in place, and ensure that the contract has appropriate warranties and indemnities.
TRADE MARKS AND PASSING OFF
In the UK, copyright protects a “work” as defined in the Copyright, Designs and Patents Act 1988, not any individual person depicted. If a deepfake replicates a voice or likeness without copying a protected work (such as a song), copyright will not apply.
English law does not recognise a freestanding personality right. Instead, individuals must rely on a patchwork of rights, including intellectual property, passing off, confidential information, data protection and defamation.
Passing off may assist where a deepfake creates a false impression of endorsement. However, it requires proof of goodwill, misrepresentation and damage, and is most useful where the individual has established goodwill or reputation (and is a known figure).
Trade mark law helps as part of the strategy where a name, image or other indicia functions as a badge of origin in trade. However, its application to deepfakes is limited. The requirement for use of a trade mark in the course of trade in order to establish trade mark infringement means non-commercial deepfakes are unlikely to be caught by trade mark law.
PERFORMERS’ RIGHTS AND MORAL RIGHTS
Performers’ rights may apply where recordings of performances are used or reproduced in AI systems.
Moral rights, including rights of attribution (the right to be identified as the author when the work is used publicly) and integrity (the right to object to derogatory treatment of the work) may apply where AI outputs misattribute or distort works, although their application to AI-generated content is uncertain.
PRACTICAL ACTION POINTS
- Audit AI use – understand how your content and likeness is being used in AI systems. Carry out due diligence on how AI is used by any suppliers or software you use.
- Strengthen contracts – review contracts including licensing agreements, collaboration agreements, media production and influencer contracts. Include clear provisions on ownership of IP, AI training, synthetic media and use of voice, image and likeness. Include an audit right and contractual protections such as warranties and indemnities.
- If deepfakes use your personal data (e.g. biometric data) without your explicit consent, you may be able to pursue remedies under the UK GDPR and the Data Protection Act 2018.
- Document clear consent – ensure express permission for any AI-related use of creative output.
- Use IP rights strategically – where applicable, consider copyright, performers’ rights, trade marks, moral rights and passing off.
- If the deepfake results in reputational damage, consider a defamation claim.
- Monitor online platforms and submit takedown requests where appropriate.
- Prepare for enforcement – document evidence of infringement from early stages.
For further information, please contact the IP & Commercial team.
This article is for general information purposes only and does not constitute legal advice or a comprehensive statement of the law. Specific legal advice should always be sought in relation to individual circumstances.
Meet the team:
Maya de Silva
- Associate
- IP & Commercial
