The significance of this trend is not confined to the entertainment industry. A recent example occurred during the Irish presidential election last year. An AI-generated deepfake news report falsely claimed that election candidate, and eventual winner, Catherine Connolly withdrew her candidacy before the vote took place. As AI-generated media becomes more sophisticated, businesses are confronting technologies capable of reproducing aspects of human identity with increasing realism. Voice, cadence, body language and biometric identifiers are beginning to emerge as commercially sensitive assets within AI-enabled markets.
Broadcasters, musicians, sportspeople and public figures are all beginning to explore how far existing IP systems can stretch when human identity itself has become commercially replicable. Examples are sports celebrities Luke Littler and Cole Palmer recently applying to file a passport photo of their face as a trade mark to seek some protection for their image. Those applications are awaiting the determination of the EUIPO on whether face images are distinctive to qualify for trade mark protection.
The EUIPO determination relates to an earlier case concerning Dutch fashion model, Maartje Verhoef. All of these applications are considered necessary to combat deepfakes as EU countries don’t recognise a personality right like the US does. Often a legal action would have to come under unfair competition laws like passing off which can be difficult to establish. Whether these applications ultimately succeed may be less important than the broader signal they sent.
Other recent filings and disputes illustrate the point. TAS Rights Management LLC is the entity managing Taylor Swift’s intellectual property portfolio. They filed three new applications before the USPTO in April 2026. These included, two sound marks seeking protection for spoken phrases including “Hey, it’s Taylor Swift” and “Hey, it’s Taylor.” The applications appear to represent an attempt to anchor recognisable vocal identifiers within a formal trade mark framework.
Under US trade mark law, sound marks like Taylor Swift’s can be registered where they function as indicators of commercial origin. The TAS Rights Management LLC applications reportedly rely on longstanding commercial use connected with entertainment and streaming services. More broadly, US law recognises both privacy and publicity rights. These rights allow individuals to commercially exploit, license and protect aspects of their persona. In some jurisdictions, these protections continue even after death.
The European position is more complicated, as human identity traits do not sit comfortably within either traditional national or EU trade mark systems. Under Article 7(1)(b) of the EU Trade Mark Regulation, a sign must possess sufficient distinctive character to function as a trade mark. In practice, the EUIPO and national EU intellectual property offices apply a relatively high threshold to non-traditional marks. This is particularly the case where a sign may be perceived, rather than a badge of origin, as:
- Commonplace speech
- Promotional language, or
- Part of an artistic performance.
This high threshold creates an immediate difficulty for voice-based applications. A recognisable voice may be commercially valuable without necessarily functioning as a trade mark in the European sense.
Digital replicas
The position becomes more complex when realistic digital replicas or photorealistic avatars are involved. Article 7(1)(e)(iii) EUTMR prevents registration of signs consisting exclusively of characteristics which give substantial value to the goods themselves. The provision has not yet been fully tested in the context of AI-generated likenesses. However, it raises interesting questions for highly realistic digital twins. Ironically, the more commercially valuable and realistic a replica becomes, the more likely it may be to encounter objections under the existing framework.
There is also a significant structural dilemma because trade mark law generally treats registered signs as fixed commercial identifiers. In contrast, living human identity is dynamic. People change appearance over time. European trade mark law has historically struggled with subject matter that changes naturally over time or cannot easily be represented in a stable and precise manner. Olfactory marks, which consist of distinctive smells or scents, have historically been among the most difficult trademarks to register. The challenges associated with representing a scent in a clear and precise manner have limited their practical use. These difficulties raise a broader question as to whether static representations of human likeness are fully compatible with living, evolving identity.
Last year, H&M partnered with Uncut to create AI “twins” of 30 real-life models for use in marketing and social media campaigns. Industry standards for these arrangements remain inconsistent. However, these projects typically involve:
- Extensive scanning
- Voice recording, and
- Movement capture processes.
H&M publicly stated that participating models would continue to receive compensation for the use of their digital replicas. The models will also retain some degree of control over how their likenesses are used, according to the fashion retailer.
AI-generated twins may prove more successful in satisfying technical requirements relating to ongoing commercial use. However, AI systems are becoming increasingly capable of generating adaptive and continuously updated replicas. As a result, future disputes may involve a more digital forensic assessment of similarity, rather than solely relying on traditional human comparison alone. Existing intellectual property systems were simply not designed to manage this tension between static registration systems and biological identity.
Elsewhere, disputes involving personality rights, digital likenesses and synthetic impersonation are becoming increasingly common. From Brooklyn Beckham to Jo Malone, individuals are beginning to publicly question the extent to which intellectual property systems should be capable of exerting long-term commercial control over aspects of personal identity. See our recent article on Estée Lauder suing Jo Malone.
Trade marks and AI
Another source of legal complexity is that trade mark was never designed to regulate the training phase of AI systems. The ingestion of photographs, recordings, voice samples and biometric material during model training typically occurs within internal technical infrastructure rather than in a consumer-facing commercial setting. Trade mark infringement traditionally requires “use in the course of trade” capable of affecting origin functions or creating consumer confusion. AI training activity therefore does not fit neatly within conventional infringement analysis.
The stronger legal leverage only emerges downstream, when generated outputs are deployed commercially. This is particularly the case where their use implies endorsement, affiliation, sponsorship or authorised collaboration. Defensive registrations may still prove strategically useful in these scenarios. Defensive registrations strengthening enforcement positions against misleading synthetic endorsements, cloned brand ambassadors or commercial impersonation. Conventional legal principles like passing off have assisted celebrities in pursuing a protection akin to a personality right when it amounts to unfair competition.
Copyright law has also been considered as a potential alternative means of protection. For example, Denmark has suggested that copyright law could serve as an appropriate vehicle for a personality-type right. Under this approach, a person’s appearance and voice would receive copyright protection. This could help prevent the creation and use of AI-generated copies. Rather, it would create a statutory control right aimed at realistic digital imitations of identifiable personal characteristics. The key operative part of the proposal is consent from the person imitated before the deepfake content is made available. It remains to be seen whether this proposal will gain traction at the national level across the EU. It is also unclear whether it could eventually develop into an EU-wide initiative.
The fast-moving and often unpredictable ways in which AI systems generate commercially valuable identities also deserve reflection. A viral trend involving so-called “Italian Brainrot” content spread rapidly during 2025 across social media platforms. The trend centred around AI-generated hybrid characters paired with dramatic synthesized narration. One of the more notable legal issues arising from the phenomenon was the origin of the characters. They did not emerge from any clearly identifiable human creator or coordinated commercial strategy. Instead, they evolved through decentralised user participation and iterative AI-generated content. This created uncertainty around authorship, ownership and enforceable rights.
Attempts to commercialise this trend quickly emerged under the EU’s first-to-file trade mark framework. Multiple applications from different applicants and multiple oppositions quickly followed. The EUIPO recently reopened examination of certain filings and issued provisional refusals under Article 7(1)(b) EUTMR. This followed the submission of third-party observations contending that terms used to describe broader cultural trends should remain available for general use and should not be monopolised through trade mark registration.. This social media phenomenon demonstrates that even impersonal or collaboratively generated digital identities may generate substantial short-term commercial value and quickly become commercially contested territory under first-to-file trade mark systems.
Taylor Swift’s IP strategy is interesting not simply as a playbook for celebrity rights, as it reveals a broader transformation taking place across intellectual property law. Businesses are increasingly attempting to treat elements of identity, including voice, likeness, behavioural patterns and biometric signals, as commercially protectable assets within AI-enabled markets. This trend is exposing the limits of legal systems built for a very different technological environment. It remains unclear how far existing trade mark and copyright frameworks can or should stretch to accommodate synthetic identity disputes. Many of the most difficult questions remain largely untested, particularly within Europe.
AI due diligence can no longer be separated from intellectual property due diligence, privacy analysis or identity risk. In an economy increasingly shaped by AI-generated content and automated systems, the greatest commercial risk may be found in the training data that powers them.
Practical steps for companies
For many organisations, the challenge is no longer simply protecting intellectual property in the traditional sense. Instead, it is understanding how identity, data provenance, contractual control and AI governance increasingly intersect. There are several practical steps businesses, rights holders and investors can begin taking:
- Audit existing intellectual property portfolios. Identify whether valuable identity assets, are adequately protected, including:
- Voices
- Signatures
- Likenesses
- Slogans, or
- Digital brand elements
- Review any relevant contracts involving models, influencers, employees or contractors. Determine whether AI training, synthetic replication and digital twin rights have been clearly addressed.
- Assess whether datasets, training materials and biometric inputs used within AI systems can be clearly traced and lawfully justified from both an intellectual property and GDPR perspective.
- Examine third-party AI licensing arrangements carefully. In particular, where foundation model providers impose restrictions relating to commercial deployment, territorial use or change of control.
- Develop internal governance procedures around synthetic media outputs, impersonation risks and AI generated content review processes before disputes arise.
- Consider whether AI related diligence should now form part of standard investment, acquisition and procurement exercises, including specialist technical review.
For more information and expert advice, contact a member of our award-winning Intellectual Property team.
The content of this article is provided for information purposes only and does not constitute legal or other advice.
