Though it has many legitimate uses, the dark side of deepfake technology has once again been highlighted by Labour MP Jess Asato’s current claim against Elon Musk’s xAI for its Grok feature which produced fake sexualised images of her.

Such non-consensual sexual deepfakes entail highly gendered harms that primarily harm individual targets, as well as women and LGBTQ+ groups more generally.  For this reason, requesting, creating, sharing and threatening to share such deepfake content is now criminalised (in sections 66B, 66E-G of the Sexual Offences Act 2003, as amended).

Jess Asato’s civil claim is based on misuse of private information (MPI) and data protection.  This post looks at the MPI aspect of her claim, arguing that the tort has the potential be incrementally developed to protect the targets of non-consensual intimate deepfakes.  In doing so, it provides an overview of more detailed arguments I have set out in a piece (‘Saving Souls’), which will be featured in a forthcoming edited collection, ‘Deepfakes & the Law’ (Routledge) that I have co-edited with Tom Bennett.

Though deepfake technology is relatively new and Asato’s is the first reported claim to come before the English courts, MPI has protected the Art 8 privacy rights of photographed individuals from its very inception.  The founding cases of Campbell v MGN (2004) and Douglas v Hello! (2005) both held that there had been Art 8 violations by the publication of intrusive photographs.  Furthermore, non-consensual sexual deepfakes occur against the backdrop of two very well-established legal principles.  First, as the European Court of Human Rights has repeatedly confirmed, most notably in Dudgeon v UK (1981), the Art 8 right protects sexuality as a ‘most intimate aspect of private life’.  Second, there is a solid line of MPI cases – including Mosley v News Group (2008), Contostavlos v Mendahun (2012) and AMP v Persons Unknown (2011) – where the English courts have protected the cogent Art 8 rights of individuals featured in intimate sexual images or footage.  This is bolstered by Supreme Court confirmation (in ZXC v Bloomberg (2022)) that individuals will usually have a reasonable expectation of privacy in relation to their nude body and details of intimate relationships.

Misuse of private information, falsehood and images

The preceding cases clearly indicate that MPI is potentially applicable to non-consensual explicit deepfakes.  But one key point distinguishes earlier MPI cases such as Contostavlos from disputes concerning synthetic deepfake footage or imagery: traditional photographs or video reveal events that actually occurred, whereas deepfakes depict the targeted individual in an entirely fictional situation.  Nevertheless, existing MPI principles offer a sound basis to resolve this disparity.  It is well established in doctrine (and confirmed by the Supreme Court in ZXC) that MPI can protect the individual’s Art 8 right whether the disputed information is true or false.  This point is demonstrated in a number of cases involving contested literary depictions of claimants.  For example, in McKennitt v Ash (2006) and Bull v Desporte (2019) the courts held that truth/falsity was not relevant to whether the information was private, and restrained publication in both cases.  And in P v Quigley (2008) the court granted an injunction to restrain the online publication of a novella that depicted a thinly-disguised version of the claimant couple engaging in ‘unsavoury’ sexual practices.

So if existing MPI offers individuals privacy protection from false text-based depictions, should this extend to false visual and/or audio-visual depictions?  Again, a survey of MPI case law would suggest so: photographic information has often been deemed ‘special’ and more intrusive than equivalent text-based information because it conveys events in more detail.  Furthermore, deepfakes have potentially greater capacity than textual accounts to mislead audiences into believing the depicted events actually occurred.  So, to avoid a logical gap in privacy protection, MPI should extend beyond privacy-violating text-based depictions (as in Quigley) to also cover equivalent false audio-visual depictions (which are generally deemed more intrusive, and potentially more likely to mislead audiences).

Clarifying the domestic approach to images

The courts will need to address a further doctrinal ambiguity in order to resolve the Asato v xAI dispute, specifically regarding whether MPI will expressly adopt a Continental-influenced approach which assumes a right to control one’s image.  In cases such as Campbell, Douglas v Hello! (2007) and Murray v Express (2008), the courts have repeatedly denied the existence of a sui generis image right or Continental personality right in UK law.  Yet, despite such denials, MPI has entailed a shift in how one’s image is understood, and this shift could benefit deepfake targets such as Jess Asato.  The Continental approach to image is encapsulated in the European Court of Human Right’s claim in Reklos v Greece (2009) that:

“A person’s image constitutes one of the chief attributes of his or her personality, as it reveals the person’s unique characteristics and distinguishes the person from his or her peers.  The right to protection of one’s image is thus one of the essential components of personal development and presupposes the right to control the use of that image.”

This passage – quoted in MPI photograph cases including Stoute v News Group (2023)  and Weller v Associated News (2014) – depicts one’s image as a distinct possessable object, and gives the individual a right to control it in principle.  This move towards a quasi-propertised notion of image is by no means unproblematic.  But it has the potential to benefit deepfake targets by enabling them to object to the unauthorised use of their image in intrusive sexual deepfakes.

Adjustments to the 2-stage test

In any event, to successfully establish MPI a claimant such as Jess Asato will have to first show they have a reasonable expectation of privacy, and, if so, the court will then balance their Art 8 privacy right against the defendant’s Art 10 free expression right.

Whether a claimant has a reasonable expectation of privacy is a flexible, fact-sensitive assessment guided by the factors set out in Murray v Express (2008).  Many of the Murray factors have a clear relevance to unauthorised intimate deepfakes.  For example, the absence of consent, the severely harmful effect upon the claimant and the explicit, demeaning nature of the information would all count heavily in favour of a reasonable expectation.  However other Murray factors – e.g. circumstances in which the information came into the publisher’s hands – fit more readily with the traditional tabloid mass media intrusions that shaped MPI, and are therefore more difficult to map onto new deepfake technology.

At stage 2 balancing, the courts gauge the respective values of the conflicting Arts 8 and 10 rights claimed in the particular case to decide which one should prevail.  This relatively unstructured weighting process is guided by Re S (2004), domestic MPI public interest principles and the factors set out by the European Court of Human Rights in Axel Springer v Germany (2012).  According to these existing authorities, the rights-balancing in unauthorised sexual deepfake disputes should be relatively clear and straight forward.  As explained above, claimants are deemed to have very strong Art 8 rights in relation to intimate photographs and footage, and the claimant’s position as a ‘public figure’ does not affect this.  In contrast, the Art 10 rights of defendants in relation to such material are generally afforded low value (as in Mosley and AMP).  Furthermore, commercial-value-based Art 10 arguments of that sort that X might advance in its defence have been routinely outweighed when previously argued by tabloid media defendants to justify intrusive publications.  In short, non-consensual sexual deepfakes cannot be justified on established stage 2 principles, e.g. that they contribute to a debate of general interest.  Looking to broader Art 10 ECtHR principles, nor could they be credibly deemed artistic or satirical expression, particularly where they mislead or confuse audiences as to whether the depicted events occurred.

Ultimately, if MPI is to remain relevant in the digital age and avoid becoming a relic, it should continue to develop in in order to meet the new challenges that digital technology poses to individuals’ Art 8 privacy rights, whilst also accounting for a wider array of legitimate expression by net-users, such as artistic and satirical deepfakes.  With these modest adjustments to the two-stage test (detailed further in my chapter), MPI could protect the targets of unauthorised sexual deepfakes.  This is not to suggest that MPI is a panacea, or that no additional regulation is needed to more effectively address pernicious deepfakes or gendered online harms.  But existing private law should urgently protect the targets of non-consensual intimate deepfakes and, with some modest adjustments, MPI is up to this vital task.

Rebecca Moosavian is Associate Professor in Law at Leeds University.