Consider this hypothetical scenario: You are a famous actor, you’ve played the role of a merchant mariner taken hostage by pirates, a pilot and a character who likes to run and whose mama told them life was like a box of chocolates. One day, you wake up to see a hyper-realistic video of yourself on the internet advertising a dental plan that you’re unaware of.
Except it’s not a hypothetical scenario, as artificial intelligence (AI) was used to generate a deepfake video of Tom Hanks promoting services that he has no affiliation to.[1] This example shows how deepfakes can exploit a person’s image to falsely endorse products without their consent. The question is, what is a deepfake and is there any legal recourse in the UK for Tom?
What is a deepfake?
In simplistic terms, a deepfake is a ‘fake’ image, audio, or video clip that has been generated by AI that uses ‘deep-learning’.[2] The AI uses existing images, audio files and videos that have been uploaded to the internet to piece together a person’s likeness.
Unfortunately, deepfakes are often used maliciously, as exemplified in the recent media coverage concerning the AI-generated images of Taylor Swift.[3] Additionally, there is concern that deepfakes could be deployed in political elections to spread misinformation. In the business context, deepfakes are becoming a common tool for committing fraud. For example, a UK-based energy company was scammed out of roughly €220,000 when an employee believed they were on the phone with the company CEO who instructed them to transfer funds to a supplier.[4]
Despite the above, deepfakes can also be used for good. For example, there is a company that is using voice cloning to help people with ALS (Amyotrophic lateral sclerosis) protect their ability to speak.[5]
What is the UK’s legal position on deepfakes?
On 16 April 2024, the UK government announced that creating sexually explicit deepfakes will become a criminal offence.[6] However, at present, no legislation in the UK specifically deals with deepfakes, nor is there a specific right that protects a person’s image as reiterated by Mr Justice Birss in Fenty v Arcadia Group Brands Ltd.[7]
There are existing laws, for example, data and privacy laws, which could potentially be applied to certain circumstances involving deepfakes, however, the applicability of these laws will depend heavily on the facts of the case.
From an intellectual property law perspective, an obvious course of action would appear to be a claim in either copyright or passing off. However, such claims may not be so straightforward due to the unique nature of deepfakes and the process involved in their generation.
Copyright and deepfakes
Copyright law is concerned with protecting images and expressions of ideas. So, perhaps a copyright infringement claim could offer recourse to Tom. However, rarely does the ‘victim’ of a deepfake own the copyright in all the images that have been used to train the AI. This makes it difficult to bring an infringement claim. Furthermore, there is an ongoing debate in the UK as to whether the use of copyright material to train AI constitutes an infringement in the first place. Even if the victim owned all the copyright and there was a valid infringement claim, the deepfake creator might still get away with it by relying on a legislative exception; for example, fair dealing by claiming that the deepfake was intended to be a parody.[8]
Passing off and deepfakes
The law of passing off has expanded in scope over the years to cover more than just the classic situation where a seller misrepresents to a buyer that their goods are the goods of another. Therefore, if Tom can 1) establish goodwill, 2) show that non-consensual use of his image misrepresents the fact that he endorsed certain goods or services, and 3) prove that his goodwill and reputation will consequently suffer, it’s plausible that a passing off action might be successful (subject to the facts of course).
In a somewhat analogous situation, the law of passing off was successfully used to protect a celebrity in a false endorsement case, namely, Irvine v Talksport.[9] In this case, Talksport was using a photo of Eddie Irvine in an advert without his permission, suggesting that he endorsed their radio station. It was successfully argued that Eddie was well enough known to a significant portion of the relevant public to have substantial goodwill [10] in his name and likeness to support his passing off claim.
Where a person doesn’t have a significant media presence, as would probably be the case with the aforementioned energy company CEO, it will be difficult to establish the requisite level of goodwill. Consequently, a passing off claim in these circumstances would probably fall at the first hurdle.
However, assuming that a celebrity can establish goodwill (as Eddie Irvine did), the argument in a deepfake context is then likely to turn on the misrepresentation point, in other words, is the deepfake believable enough to deceive the public into thinking that the celebrity is genuinely endorsing a particular product. If the deepfake is of poor quality or obviously intended to be amusing, consumers are unlikely to believe that there is a genuine association and therefore there cannot be any damage resulting from a misrepresentation.
This brings us to the third requirement in the ‘classical trinity’ of passing off, namely, damage. In this context, the claimant, Tom, would likely be looking to prove damage to reputation or potentially loss of endorsement revenue (if he was considering branching out into dentistry advertisement). One could argue that his reputation may be jeopardised by the fact that he no longer has control over the use of his image. However, bearing in mind Birss J’s comments regarding the lack of an image/personality right, the courts may be unwilling to accept this.
Therefore, in theory, it seems logical to apply passing off law to a deepfake false endorsement case in light of the decisions in Irvine v Talksport and Fenty v Arcadia. However, until a case like Tom’s is argued in the UK courts, we can only speculate as to how the law will be applied in such circumstances.
Conclusion
In conclusion, in the UK, legal responses to deepfakes are still evolving. Despite recent criminalization efforts for certain deepfake uses, no specific legislation addresses the broader implications as exemplified by the case of Tom Hanks’ unwitting endorsement. Continued vigilance and collaboration among lawmakers, legal experts and society are crucial to effectively address the challenges posed by deepfake technology, while leveraging its potential for positive innovation.
[1] Tom Singleton, ‘Tom Hanks warns dental plan ad image is AI fake’ (BBC, 2 October 2023) <https://www.bbc.com/news/technology-66983194.amp> accessed 17 April 2024
[2] <https://www.oed.com/dictionary/deepfake_n?tab=meaning_and_use#1345352340> accessed 17 April 2024
[3] Imran Rahman-Jones, ‘Taylor Swift deepfakes spark calls in Congress for new legislation’ (BBC, 27 January 2024) <https://www.bbc.co.uk/news/technology-68110476> accessed 3 May 2024.
[4] Jesse Damiani ‘A voice deepfake was used to scam a CEO out of $243,000’ (Forbes, 3 September 2019) <https://www.forbes.com/sites/jessedamiani/2019/09/03/a-voice-deepfake-was-used-to-scam-a-ceo-out-of-243000/> accessed 17 April 2024
[5] <https://www.projectrevoice.org/> accessed 17 April 2024
[6] <https://www.gov.uk/government/news/government-cracks-down-on-deepfakes-creation> accessed 17 April 2024
[7] [2013] EWHC 2310 (Ch) [2] (Birss J)
[8] Copyright, Designs and Patents Act 1988, s30A
[9] [2002] EWHC 367 (Ch)
[10] Above n 9, [46] – [47] (Laddie J)
