Most recently, Taylor Swift has been swept into scam advertisements on TikTok, where AI-generated replicas of her voice and likeness are used to promote fraudulent reward schemes.
In response, she registered two sound marks with the U.S. Patent and Trademark Office (“Hey it’s Taylor Swift…” and “Hey it’s Taylor…“) and one image mark (of her playing a pink guitar on the Eras tour) to protect her likeness and voice against deepfakes. Earlier this year, actor Matthew McConaughey took a similar approach, pursuing protection over elements of his image, video presence and distinctive vocal expressions, including his iconic “alright, alright, alright” catchphrase from “Dazed and Confused”.
Artists and celebrities registering their likenesses is nothing new; a photo of Tom Jones was registered as a trade mark in the UK and the EU over 20 years ago. Since a radio was digitally placed into Eddie Irvine’s hand in Irvine v Talksport[1], AI technology has only become more widespread, sophisticated, and increasingly difficult to detect.
The trend continues across entertainment and sport. Darts champion Luke Littler registered his face as a UK trade mark, to combat AI fakes and footballer Cole Palmer has registered his autograph, facial image, and the phrase “Cold Palmer” – after his well-known ‘shivering’ celebration.
The question now is whether existing IP rights are fit for purpose to prevent unauthorised use.
Questions surrounding the protection of likenesses are being tested before European authorities. The EUIPO Grand Board of Appeal is currently considering whether the representation of a person’s face (Johannes Hendricus Maria Smit, a Dutch TV presenter) is eligible for registration as a trade mark under Article 7(1)(b) and 7(1)(c) EUTMR. This cases raises important questions about distinctiveness and whether consumers would perceive a face alone as an indicator of commercial origin. Read our previous article on this case here.
While trade mark protection is a vital component of overall brand protection, there are significant hurdles when securing registration of trademarks, which can often be a lengthy and sometimes uncertain process, particularly where objections or oppositions are raised for proposed applications. Registration also does not confer an absolute monopoly over a mark, let alone a brand. Protection is limited by the scope of the registration (i.e. the classes the mark is registered under), and the extent to which third-party use creates a likelihood of confusion. A registered trade mark proprietor would also need to show that any offending use qualifies as “infringing use” – i.e. the use is i) in the course of trade, ii) in relation to goods and services, iii) there is no consent from the trade mark proprietor and iv) the use affects, or is liable to affect the functions of the trade mark.
Critically, trade mark rights are also territorial in nature, meaning protection is generally confined to the jurisdiction where the mark has been registered. For unregistered marks, the process of obtaining registration typically presents as the biggest barrier to brand protection.
Further, even if successfully registered, there remain additional pitfalls, and personalities should ensure that IP ownership is structured in an appropriate way. In 2008, footballer Diego Maradona registered his name as an EU trade mark, covering a wide range of goods and services, including clothing and footwear. When he passed away in 2020, his former lawyer’s company, Sattvica S.A., attempted to claim ownership of the trade mark portfolio and sought to enforce rights in the Maradona name – even against Maradona’s own children who had been posting the footballer’s name and likeness on social media. Whilst registration of a notable name can strengthen protection, the risk of third-party oppositions can significantly complicate the exercise of enforcement of those rights.
Some additional rights protect personality rights, however they are limited, and depend on the territory in question. Under English law, there is no specified image or personality right, unlike in the US, so there is arguably a legal ‘gap’ in how these issues are tackled. In Fenty v Arcadia Group Brands Ltd,[2] the UK Court of Appeal held that consumers could be misled into thinking that Rihanna had endorsed a T-shirt with her face in it, and so Topshop’s sale of those unauthorised t-shirts was found to be passing off Rihanna’s goodwill. Passing off can offer some protection but is limited. A successful passing off claim usually requires the personality to have generated goodwill in their personal brand, so they must have already lent their name/likeness to brand endorsement (Irvine v Talksport). Further, a passing off action is significantly more time and resource consuming than a trade mark infringement action, as claimants must demonstrate clear evidence of their goodwill.
It is therefore important that artists have a clear brand protection strategy in place. They should first pay attention to how their name and personal brand is protected, and ensure core trade marks are registered, and held by the correct entity. Then, more unconventional trade marks, such as voices and likenesses, can be considered – understanding that these trade marks have their limitations.
Beyond trade mark and passing off, other protection may be offered by other legal rights, such as copyright, data protection, and privacy. Deepfakes typically involve the reproduction of copyrighted works, whether it is images or videos, which is an act of copyright infringement under s.17 of the Copyright Designs and Patents Act 1988 (CDPA).
However, there may be more challenges if seeking to rely on copyright protection. For example, many deepfakes are generated from training data scraped from across the internet which may not be copyrightable. Where a deepfake does not actually copy an existing copyright work, the protection given to the rightsholder is much weaker.
In the UK, s.180 CDPA protects certain rights of a performer, including their recording rights and exploitation of a live performance. In a deepfake context however, the performance in question is entirely AI-generated and it would be difficult to argue that the performers’ rights have been infringed at all as it’s merely a convincing imitation. In this respect, the scope of performers’ rights protection under the CDPA is limited.
The laws of defamation, privacy, and data protection are also relevant considerations when seeking to enforce rights against deepfakes, particularly where individuals appear to be engaging in gross misconduct or where they are created without the individual’s express consent.
Even where a personality has valid enforceable rights, there remains the cost and practicality of enforcement. Much infringement takes place online, and so determining the identity of the bad actor isn’t always straightforward, infringements pop up again in other locations, and often enforcement can feel like a long game of whack-a-mole. Further, access to more time and cost effective online take down mechanisms isn’t readily available without registered rights (rather than unregistered rights, like passing off).
Protection of more unusual marks/personality likeness as trade marks are unlikely to completely solve the problem of AI generated deepfakes:
Find out more about our expertise in Brand Protection here.
[1] [2003] EWCA Civ 423
[2] [2015] EWCA Civ 3
