Deepfake Sextortion in England, Wales and Northern Ireland: A Doctrinal and Regulatory Analysis

Abstract

Existing law provides no settled account of how deepfake sextortion should be characterised and regulated in England, Wales and Northern Ireland, creating uncertainty for charging, adjudication and platform compliance at the point when the Online Safety Act 2023 allocates duties to regulated services under Ofcom oversight. This article responds by analysing and synthesising the Online Safety Act 2023 with the Sexual Offences Act 2003 and residual harassment and communications offences, using doctrinal analysis and normative evaluation to identify points of alignment and misfit. It establishes criteria for identifying synthetic sexual coercion, including the elements that mark threat-stage conduct, the role of fabrication in the wrong, and the conditions under which epistemic harms should be treated as legally relevant within ordinary doctrine. It rejects three propositions: that intimate-image abuse is primarily a publication-based wrong; that an authentic image is a precondition for liability; and that content-led platform duties adequately address coercion before dissemination. This analysis specifies how courts and prosecutors should classify conduct and select offences, how services should operationalise risk assessment and mitigation for threat-stage harms, and which targeted reforms to offence design, platform duties and victim-facing procedures are required to secure predictable protection and effective redress.

1. Introduction: Deepfake Sextortion and Synthetic Coercion in UK Law

Deepfake-enabled sextortion marks a qualitative shift in cyber-facilitated gender-based violence. It combines the mimetic capacities of generative AI with the coercive logic of blackmail, using fabricated sexual imagery to induce compliance, extract payment, or silence a target. Unlike earlier forms of non-consensual pornography, where the injury arises from the unauthorised circulation of genuine intimate material, deepfake sextortion operates through fabrication and credible threat. The victim is compelled not by the revelation of what has occurred, but by the prospect that others will accept as true what has never taken place. This article terms that phenomenon “synthetic coercion”: a form of pressure grounded in falsity but effective because it is socially believable.

The central legal problem addressed is that the law of England, Wales and Northern Ireland lacks a coherent doctrinal and regulatory response to deepfake sextortion as synthetic sexual harm, despite recent reforms to intimate-image offences and the enactment of the Online Safety Act 2023 (“OSA”). Core legal categories—defamation, harassment, blackmail, and image-based sexual offences—are organised around an underlying act or a genuine recording or image. Deepfakes, by contrast, are technically false while being socially and psychologically real. The harms they generate are anticipatory, relational and cumulative, structured around the threat of reputational collapse, social exclusion or renewed targeting. Protection is often predicated on “truth”, “authenticity”, or repeated conduct at precisely the point where the wrong consists of plausible falsification and single, devastating threats.

The problem is aggravated where the target is a public figure, including an elected politician. Sexualised depictions of politicians can, in some contexts, fall within political satire and artistic expression protected by Article 10 of the European Convention on Human Rights (ECHR), and the European Court of Human Rights has treated satire—by its nature involving distortion and provocation—as a form of expression that may warrant a high level of protection (Enarsson 2025), particularly where it concerns public figures1. However, it must be made clear that deepfake sextortion is analytically distinct from satire. Where fabricated sexual imagery is used as an instrument of coercion or extortion, the conduct is not best understood as participation in public debate. It is the imposition of pressure through a sexualised falsity that foreseeably interferes with Article 8 (ECHR) interests in private life, identity, and psychological integrity. The legal task is therefore not to deny that satire exists, but to develop criteria capable of distinguishing protected political expression from coercive, threat-based abuse—especially where fabricated sexual imagery is deployed to deter participation in public life or to discipline women’s presence within it.

The article advances three interrelated claims. First, deepfake sextortion should be understood as a distinct form of synthetic sexual harm, rather than a variant of existing communications or obscenity offences. The wrong lies not merely in the circulation of images but in the weaponisation of plausible fictions to strip victims of narrative control over their sexual and digital identities. Threats that never eventuate in publication, demonstrably inauthentic images, and trauma that arise at the point of credible anticipation remain at the margins of current legal understandings of injury, responsibility and consent.

Secondly, the article argues that the OSA, although positioned as the systemic response to online harm, embeds conceptual and structural misfits when applied to deepfake sextortion. The Act’s duties of care and Ofcom’s enforcement model are built around categories of illegal content and ex post content management. Deepfakes are therefore treated primarily as items to be located and removed once they appear on regulated services, rather than as instruments of threat and coercion whose harm crystallises before publication and often outside visible platforms. The interaction between the OSA and intimate-image and communications offences reinforces this misalignment by tethering protection to authenticity and dissemination.

Thirdly, drawing on feminist legal theory and scholarship on epistemic injustice, the article develops a normative framework of coercion through falsity and proposes a programme of reform for England, Wales and Northern Ireland. The core contention is that legal accountability should be reoriented away from evidentiary literalism and towards the recognition of synthetic sexual harm as a first-order violation. That shift requires both substantive reform—such as a strict-liability base offence for deepfake abuse and recalibrated platform duties directed at threat-stage harms—and procedural redesign that embeds victim agency and credibility within ordinary criminal, regulatory and civil processes.

The article advances the literature by developing the notions of synthetic coercion and synthetic sexual harm in order to capture the anticipatory, epistemic and relational dimensions of deepfake sextortion; by analysing the Online Safety Act, the Sexual Offences Act 2003, and residual communications and harassment offences, asking not only whether particular fact patterns can be made to fit existing provisions but also whether the underlying statutory concepts of harm, consent, and authenticity can accommodate synthetic, threat-based abuse; and by evaluating institutional and regulatory capacity to respond, and articulating a model of accountability that centres victim experience and institutional responsibility.

The analysis is grounded in recent experience in Northern Ireland, with the case of Cara Hunter, a serving Member of the Legislative Assembly (MLA), as a recurring reference point. Reporting has described how pornographic deepfake material circulated during an election campaign, with foreseeable consequences for personal safety, political participation, and the victim’s ability to control public-facing identity. The case also illustrates why “public figure” status does not resolve the legal question. Even where a target is a politician, sexually explicit fabrication may function as coercion rather than comment: it can operate to silence, to deter public participation, or to extract compliance by exploiting the known dynamics of misogynistic attention and reputational sanction. The case further exposes practical uncertainty about applicable offences and the limits of content-centric regulatory approaches where harm is initiated through private threats, offshore dissemination, or rapid re-uploading across services.

These issues are not confined to Northern Ireland, nor to women targets, though women in public life are disproportionately exposed to sexualised synthetic abuse. Comparative signals also show that courts are increasingly asked to assess false sexualised allegations and doctored imagery directed at public figures, and do not necessarily treat political status as a general immunity for the dissemination of fabrications presented as fact. The point is not to collapse defamation, political speech, and sextortion into a single category; it is to clarify that a freedom-of-expression frame cannot do the analytical work when the operative mechanism is coercion through falsity.

The article proceeds as follows. Section 2 develops the conceptual framework of synthetic sexual harm and coercion through falsity, rethinking harm and consent beyond the binary of real and fake images and drawing on feminist and epistemic injustice scholarship. Section 3 turns to victim experience, using Hunter’s case and related material to map the structure of synthetic re-victimisation, victim disbelief and institutional silence. Section 4 analyses the principal doctrinal routes in England, Wales and Northern Ireland—sexual offences, harassment and communications offences—to identify the constituent elements that enable or foreclose liability in threat-stage cases. Section 5 draws together the preceding analysis to state the implications for enforcement and compliance and to set out a reform programme framed in terms of offence structure, calibrated platform duties and victim-centred procedural mechanisms.

2. Conceptual Framework: Rethinking Harm and Consent in the Age of Deepfakes

At the core of deepfake sextortion lies a jurisprudential difficulty: how should law conceptualise harm, coercion and consent when the material at issue is fictitious, yet the consequences are acutely real? The dominant architecture of sexual offences and cognate doctrines ordinarily links sexual harm either to corporeal violation or to the unauthorised disclosure of factually accurate intimate information. Deepfake sextortion does not fit this template. There is no physical contact, no antecedent sexual encounter and no breach of factual privacy. The injury instead arises from the threatened manipulation of identity through synthetic representation: what is put at risk is not the integrity of the body, but the stability of the person’s social persona and standing.

This section develops a conceptual account of deepfake sextortion as a form of harm that unsettles established legal categories and requires doctrinal and theoretical revision. It argues that synthetic, threat-based sexual abuse exposes the limits of a framework in which wrongful conduct is premised on material participation or factual disclosure, and that a more adequate model must treat the coercive deployment of plausible fabrications as a distinct juridical wrong.

2.1. From Physical Violation to Representational Assault

The concept of crime has never been static. It has shifted with changes in social ordering, legal systems and moral interpretation. Early legal arrangements tended to correlate crime directly with manifest harm: actus reus was inferred from the concrete effects of the impugned act or omission. As criminal law developed, this causation-based conception expanded to encompass indictable offences not because of demonstrable consequences, but because the prohibited conduct is deemed wrongful in itself. In such cases, proof of the act alone may suffice for conviction, irrespective of any measurable downstream harm. Deepfake sextortion both draws on and disrupts this trajectory. It is grounded in synthetically generated acts rather than physically enacted ones, yet it produces harms that are socially and psychologically real.

A preliminary difficulty is whether the harms described in this section should be treated primarily as matters for civil law rather than the criminal law. Liberal accounts of criminalisation typically require a clear justificatory basis for state punishment and have often treated reputational or representational injury as falling first within civil remedies, particularly defamation. This allocation is partly explained by the need to preserve a margin for error, exaggeration and criticism in public discourse—what Weaver et al. describe as a qualified “right to be wrong”—a consideration that applies with particular force where the claimant is a politician or other public figure (Weaver et al. 2006). The same concerns have historically informed scepticism about criminalising reputational wrongs (including criminal libel) and proposals for its reform or retrenchment (Walker 2005–2006). These points do not preclude criminal liability where additional elements are present (such as coercion or extortion), but they require the analysis to distinguish deepfake sextortion from mere false depiction or robust political comment.

The sexual offences law rests on an assumption of corporeal trespass. The body, and its inviolability, functions as the primary locus of harm. Harm is conventionally framed in terms of physical contact of a sexual nature without the victim’s consent. Subsequent extensions of the criminal law to image-based sexual abuse—such as cyberflashing and “revenge pornography”—remain connected on this logic.2 They presuppose authentic material documenting real bodily exposure or intimacy and criminalise the non-consensual dissemination of those genuine intimate images. The focus lies not on the creation of such images, nor on their source, but on their circulation. Empirical work suggests that victims are frequently attributed partial blame for having shared intimate images with a partner, thereby diluting, to some extent, the perceived weight of their non-consent to later dissemination (Gavin and Scott 2019; see also Starr and Lavis 2018). Although such assumptions rest on normative stereotypes, the underlying image is not in dispute; the controversy attaches to its distribution (Donevan 2025).

Deepfake sextortion effects a fundamental shift. It presents a synthetic threat that generates real harm. The images are fabricated and distorted; the harm materialises not only, or even primarily, at the point of distribution but at the moment of creation and threat. Deepfake sextortion thus produces a novel category of wrongdoing grounded in representational injury, threat based on synthetic image generation and real social harm in the absence of any prior sexual act or breach of factual privacy. The injury is constituted by the expropriation of a person’s digital likeness and the imposition of a false sexual narrative under the threat of exposure. In challenging the assumption that sexual harm presupposes bodily violation or unsolicited exposure of genuine intimacy, deepfake sextortion destabilises the theoretical foundations of sexual offences jurisprudence and calls for a reconceptualisation of actus reus in the domain of sexualised abuse.

A further clarification is required. Representational injury is not unique to sexual depiction; defamation law addresses a wide range of false imputations capable of lowering standing or causing distress. Nor does the analysis depend on a claim that sexual depiction is inherently “worse” than depictions of corruption, child abuse, plagiarism, or other serious misconduct. The argument is narrower: deepfake sextortion is distinguished by the mechanism of wrongdoing—coercion through a credible sexual falsity—and by the way sexualised fabrication is predictably used as leverage to induce compliance, silence or payment. This mechanism often brings the conduct within, or adjacent to, the law’s established concern with coercion, extortion and targeted abuse, rather than leaving it to the civil management of reputational dispute. The analysis below therefore treats “sexual” not as a universal hierarchy of harm, but as a recurrent and effective modality of coercion with doctrinal consequences for actus reus, temporality (threat-stage harms), and attribution of responsibility

Feminist theorists have long questioned the law’s fixation on physicality (Donevan 2025). MacKinnon argues that sexual injury is frequently realised through symbolic domination and the production of inequality via sexual representation, rather than through direct bodily intrusion (MacKinnon 1993). For MacKinnon, sexuality is a pervasive dimension of social reality structured by dominance and subordination. The “scenario of sexual abuse”, as she famously observes, is captured in the simple imperative: “you do what I say”(MacKinnon 1989). From this vantage point, the statutory focus on force and consent in rape law normalises coercion to the extent that it is treated as criminal only in the absence of permission. Force, in letter and application, is not per se wrongful; it becomes so only when consent is withheld. This construction risks perpetuating the idea that women may “enjoy” forcible penetration within a framework of male supremacy and thereby dilutes the conception of rape as an affront to sexual integrity. What matters, on MacKinnon’s account, is not merely whether force or coercion is present, but the way in which sexual relations are embedded in objectification and the consequent subordination of women. Through the masculine lens she describes, “sex” is not clearly distinguishable from violence; desire is extracted from inequality, debasement, humiliation and intrusion in order to secure what is ultimately sought: control (MacKinnon 1989).

Nussbaum, while critical of certain strands of feminist theory, approaches these issues through a liberal-humanist lens, advocating a capabilities-based account of dignity and equality (Nussbaum 1999). She argues that the appropriate response to violence against women is not the abandonment of liberal ideals but their more rigorous realisation. Liberalism, in her view, can secure “equality of capabilities” only if resources and institutional arrangements genuinely empower individuals to live in accordance with their own values and choices. On this account, dignity is articulated through a set of central human capabilities, and a failure of social justice occurs when a society cannot secure these capabilities for all. Of particular relevance is the capacity for affiliation, which entails self-respect and protection against humiliation—the capacity to be treated as a dignified human being among others.

Nussbaum’s critique thus reframes the problem: the deficit lies less in liberal ideals of personhood, dignity and self-respect than in their weak institutionalisation and enforcement (Nussbaum 1999). When these feminist insights are brought to bear on deepfake sextortion, the phenomenon can be understood as a paradigmatic instance of representational assault. As MacKinnon’s analysis suggests, sexual harm may manifest through symbolic practices that encode dominance and subordination (MacKinnon 1993). In deepfake sextortion, algorithmically generated sexual images pose a real threat to bodily autonomy, reputation and dignity by constructing a false appearance of sexual availability and holding it hostage. The threat is perceptual, however causatively real: it reorganises the victim’s social world by recalibrating how she is seen, spoken about and treated. Read through Nussbaum’s capabilities framework, this practice undermines affiliation and self-respect by subjecting the victim to humiliation and by constraining her effective freedom to participate in public and private life on equal terms (Nussbaum 1999).

Deepfake sextortion is therefore best understood as a form of discursive sexual violence: a coercive practice that appropriates and distorts visual identity in order to compel, degrade and silence3. The mechanism of control lies in the threatened circulation of a fabricated image, crafted to appear plausible to third parties and thereby to subordinate the victim’s agency. The law’s reluctance to recognise such representational assaults as cognisable harms discloses a deeper epistemic bias—a preference for injuries that are tangibly verifiable over those that are mediated, psychological or reputational, even where the latter are socially catastrophic. Victims of deepfake sextortion experience fear, stigma and enforced self-censorship; the looming prospect that a synthetic but credible visual narrative may be unleashed constrains their capacity to participate in social, professional and intimate life. To insist upon physical violation or factual disclosure as the precondition for legal recognition is, in effect, to exclude from view a category of coercive harm that is technologically enabled but experientially real. The task, therefore, is to reconceive harm and consent in a manner that acknowledges representation as a site of violation and that accommodates coercion through falsity as a distinctive mode of control.

In the sections that follow, this claim is tested against the doctrinal materials governing England, Wales and Northern Ireland, including the extent to which existing categories can distinguish coercive conduct from protected expression and from civil reputational dispute, and how threat-stage harms can be addressed without collapsing the boundary between criminal wrongs and the ordinary toleration of falsity in public life.

2.2. The Ontology of Synthetic Harm

The synthetic nature of deepfake material often produces institutional hesitation:if the image is fabricated, it is treated as falling below the threshold of legally cognisable harm. This does not require a crude view that harm tracks factual correspondence. It is better explained by two familiar moves: authenticity is treated as an evidential proxy for seriousness, and criminal intervention is treated as exceptional within liberal doctrine. A Millian harm principle can accommodate civil liability for reputational injury, including defamation, while remaining cautious about extending the criminal law absent a clear justificatory basis (Bell 2021).

In deepfake sextortion, that justificatory basis lies in the coercive deployment of a plausible sexual fabrication—typically coupled with threat or conditional demand—rather than in falsity as such.This article uses “synthetic harm” to denote injury that arises from plausible attribution and coercive use, not from truth-value. The decisive question is not whether the representation is empirically accurate, but whether it is used to impose a credible threat that constrains the target’s options. Where law is not the appropriate or sufficient response, non-legal measures, including education about realistic expectations of social life and reputational control, may also have a role. They do not, however, remove the need to identify when coercive synthetic fabrication crosses the line into actionable harm.

Zemiology emerged as a critique of positivist criminology’s narrow focus on legally codified wrongs and observable injuries (Hillyard and Tombs 2007). It redirects attention to social harms generated by criminal, non-criminal and systemic practices, thereby widening the analytical field beyond offence definitions (Flynn and Hall 2017). The relevance here is not to collapse the boundary between harm and crime, but to clarify how thresholds of recognition are set and how responsibility is allocated between criminal law, civil liability, regulation and non-legal responses. Victimology has undergone a parallel shift. Rather than treating victimhood as a static status, contemporary accounts examine the interactional and institutional practices through which victimhood is ascribed and contested. Victimisation, therefore, concerns not only what is done to a person, but also how the “victim” label is conferred, withheld or disputed in communicative practice (Garfinkel 2023). Accuracy in victim-labelling is a normative concern because misrecognition, minimisation or denial of victim status can entrench harm and impede access to protection and redress (Holstein and Miller 1990).

Within this framework, communicative exchange is not merely transmissive but constitutive: it can shape social relations and statuses (Emerson and Porte 1983). Speech not only describes events; it can participate in the social production of categories and standing (Heritage 1984). To identify someone as a “victim”, for example, is not simply to report a neutral fact but to engage in a classificatory act with practical implications for recognition, credibility and institutional response. Deepfake sextortion exploits this constitutive dimension of communication. The synthetic image functions as a discursive resource through which the target’s identity can be re-authored—as promiscuous, untrustworthy, or morally compromised—irrespective of any factual sexual conduct. The harm lies in the credible threat that this imposed narrative will be accepted and reiterated by others, and in the coercive leverage that follows from that anticipated uptake.

Austin’s speech-act theory provides a useful analytical lens for this phenomenon. For Austin, to utter certain sentences in the appropriate context is to do something: performative utterances bring about consequences merely by being spoken (Alsamhori et al. 2025). The utterance constitutes an illocutionary act which, under conditions of infelicity, generates additional perlocutionary effects such as fear, exclusion or restrictions on freedom (Austin 1975). On this view, the origin of harm does not depend on whether the underlying content is accurate, but on the authority of the speaker and the social context in which the utterance is made. A public accusation of criminality at a press conference can compromise reputation at the moment of articulation, even if subsequently disproved. By parity of reasoning, the act of threatening to circulate a deepfake is itself performative: it reconfigures the victim’s situation by making fear, constraint and compelled secrecy rational responses, regardless of whether the video is ever released.

Langton’s extension of speech-act theory to pornography further sharpens the analysis (Langton 1993). Building on MacKinnon and Austin, Langton argues that pornography can function as a kind of speech act that subordinates women by unfairly ranking them as inferior, legitimising discriminatory treatment, and depriving them of powers enjoyed by men (Langton 1993). The point is not merely that pornography reflects inequality, but that it can help constitute and entrench it. Within this frame, sexual assault is not simply a discrete harm or crime; it is part of a broader pattern of legitimised discriminatory behaviour. Deepfake sexual abuse can be situated within this constellation: it is a coordinated use of synthetic imagery and a threat to inscribe women into subordinating narratives, thereby constraining their social and political agency.

Butler’s account of performativity underscores its temporal and embodied dimensions (Butler 1990). Performativity, on Butler’s view, is not a one-off event but a series of repeated, ritualised practices whose power lies in their sedimentation over time. Gendered bodies are produced and maintained through these reiterations; they do not possess a pre-social, ontological core, but are constituted through normative scripts (Butler 1990)4. Applied to deepfakes, this suggests that synthetic sexual images work by tapping into familiar cultural performances of femininity and sexuality. The deepfake does not merely misrepresent an individual; it inserts her into an already intelligible script of sexual availability and shame. Its injurious potential lies in its capacity to circulate as recognisable “truth” within that normative frame, and to be believed as such.

Sextortion intensifies this dynamic. It weaponises the possibility of circulation rather than circulation itself. The video may never be released, yet the anticipation of exposure may be sufficient to traumatise the victim, to reorient her decisions and relationships, and to extract compliance or material value. In Austin’s terms, the threat operates as an illocutionary act whose perlocutionary effects—fear, withdrawal, payment, silence—are the very aims of the perpetrator. The law’s shortcoming here is one of ontological misrecognition. By reserving the status of “real harm” for tangible physical acts or the dissemination of factual images, it fails to register that injury can be inherent in the distortion and weaponisation of narrative itself.

Legal intervention is therefore not merely remedial but constitutive. Law can function as a corrective mechanism when its purpose is to reconfigure existing social norms, rather than simply to mirror them. Recognising synthetic harm as harm in its own right is part of that project. By articulating that the coercive manipulation of synthetic sexual imagery is a legally cognisable wrong, the law can reshape the moral understanding of such practices and signal that representational assaults stand on a par, in terms of seriousness, with more traditional forms of sexual and reputational violation (Sunstein 1996). In this sense, a jurisprudence of synthetic harm aspires not only to respond to individual cases but to drive normative change by reorienting social interpretation of what counts as sexual violence and coercion in the age of deepfakes.

2.3. Consent Without Contact: A Legal Fiction Reversed

Consent operates as a threshold concept in both sexual offences law and the regulation of intimate images. Traditionally, the absence of consent marks the line between permissible and wrongful conduct, whether in relation to sexual acts or the capture and dissemination of intimate material. The model presupposes an underlying act or artefact—an encounter, a recording, a disclosure—to which consent might be given or refused. Deepfake sextortion exposes the limits of this presupposition. In such cases, there is no prior sexual act, no authentic image and no factual event that could have formed the object of consent. The victim is coerced into acquiescing in the fictionalisation of their identity. What emerges is a form of synthetic non-consensuality: refusal is directed not at a real act, but at a fabricated one whose coercive power lies solely in its plausibility.

The point is not that criminal law lacks resources to address threats or coercion without material participation. On the contrary, blackmail and harassment offences exemplify long-established forms of liability that attach at the threat stage and can operate where the threatened allegation concerns what the victim is said to have done or said, including where that allegation is false. The difficulty is narrower: within the sexual offences doctrine and the mainstream framing of image-based abuse, the organising assumptions remain that there is an unconsented sexual act or an authentic intimate artefact whose capture or dissemination can be assessed through consent. Deepfake sextortion, therefore, requires a recalibration of the consent inquiry so that it is not treated as exhausted by participation, recording, or disclosure, but is directed to the use of plausible sexual fabrication as leverage.

This inversion unsettles the implicit legal fiction that sexual and image-based harms must be anchored in material participation. Framed in orthodox terms—“did the complainant engage in the act?” or “did they authorise the recording or disclosure?”—the doctrinal question no longer captures the wrong. For Herring, the core of sexual offending lies in the violation of sexual autonomy through unconsented physical contact (Herring 2022); cases such as R v Bowden crystallise liability around the absence of agreement to the act itself.5 Likewise, the literature on image-based sexual abuse locates the harm in the non-consensual dissemination of genuine intimate material, understood as an assault on privacy and dignity (McGlynn and Rackley 2017). Deepfake sextortion sits uneasily within these categories. The relevant inquiry is not whether the complainant engaged in the depicted conduct, but whether the representation was deployed to induce fear, submission or compliance—the locus of harm shifts from action to affect, from exposure to anticipation.

Criminal law theory reinforces the need for this shift. Hughes has argued that the decision to criminalise the violation of an individual interest functions not only as protection of that interest but as a public affirmation that the conduct constitutes a serious attack on shared values (Hughes 1963). In offences such as murder and rape, the statutory requirement of absence of consent operates as a formal recognition that coerced intrusion into bodily or sexual autonomy warrants the full machinery of criminal sanction. Deepfake sextortion replicates that level of intrusion, but through representational rather than physical means. The coercion is experienced as intensely as in many contact offences. However, a consent analysis tethered to a narrowly conceived actus reus cannot capture coercion that operates through synthetic threat. The result is a lacuna: conduct that is substantively analogous to serious sexual wrongdoing evades existing categories because it does not conform to their material assumptions.

A more satisfactory account of consent begins from its expressive character. For consent to have legal significance, it must be manifested through some communicative act capable of grounding claims of right or entitlement. According to Pufendorf, an unexpressed internal will cannot create obligations; it becomes legally significant only when conveyed through outward signs of assent (Seidler 2010). Wertheimer develops this into a relational standard: manifestations of consent matter morally and legally because they provide reliable evidence of an agent’s desires and choices, and because they give the other party a reason to believe they are entitled to proceed (Wertheimer 2003). On this view, the actus reus condition—physical contact or disclosure—may supply relevant context for assessing the state of mind. However, it is the reasonable belief in consent, grounded in communicative cues, that carries the normative weight.

Transposed to deepfake sextortion, this expressive model exposes a misfit in existing doctrine. The victim has issued no communicative token from which a perpetrator could reasonably infer permission to fabricate or weaponise synthetic sexual imagery (Citron 2019). On the contrary, where any cues are present, they are refusals, protests, and signs of distress. The wrong does not lie in the misreading of ambiguous assent, but in the unilateral imposition of a sexualised narrative in the face of explicit non-consent. The harm therefore, cannot plausibly be attributed to defective communication by the victim. It lies in the perpetrator’s decision to proceed as if consent to representational fabrication had been given, when it plainly has not (Citron 2019; Chesney and Citron 2019).

This reframing is consistent with the structure of blackmail offences. In R v Gallagher, although the threats concerned different subject matter, the court affirmed that the gravamen of blackmail lies in the fear instilled in the victim, rather than in the truth or intrinsic gravity of the matter threatened.6 The law of blackmail thus already treats coercion through threat as wrongful, irrespective of the veracity of what is to be revealed. Deepfake sextortion is continuous with this logic. The juridical question is not whether the complainant engaged in the acts portrayed, but whether a synthetic representation was intentionally used to secure compliance by menace. In this sense, deepfake sextortion is appropriately understood as a species of coercion through falsity, continuous with—but not reducible to—classical blackmail.

Recognising this continuity has two principal implications. First, consent must be reconceived not merely as a precondition to bodily intimacy or to the circulation of authentic images, but as a safeguard of both bodily and representational autonomy. A person’s entitlement not to have their sexual identity fictionalised and leveraged against them should be treated as falling within the protected core of sexual and communicative self-determination. Secondly, the law must move beyond a binary model in which consent attaches only to discrete acts or artefacts. It must articulate a more general principle: that it is wrongful to manufacture and deploy plausible sexual fabrications in the face of known refusal, for the purpose of inducing fear, submission or payment.

Normatively, this analysis supports both an expansive interpretation of existing offences and the development of targeted reforms. At a minimum, offences concerned with threats and extortion should be construed so that the absence of any underlying “real” conduct does not preclude liability where synthetic sexual representations are used to overbear autonomy. Beyond that, the conceptual category of synthetic non-consensuality justifies the creation of specific offences or parallel civil remedies that criminalise the creation and use of deepfakes for coercion, extortion, and harassment, regardless of the content’s authenticity. The focus of liability should lie on malicious intent and harmful consequences, rather than on the now-untenable assumption that sexual harm presupposes material participation.

2.4. Coercion Through Falsity

The common-law offence of blackmail has long recognised a foundational proposition: the coercive force of a threat is unaffected by its veracity (Galoob 2016). Section 21 of the Theft Act 1968 criminalises any unwarranted demand with menaces made with a view to gain or to cause loss; the test for “menaces” is objective and assessed by reference to a person of ordinary firmness. Judicial interpretation has consistently affirmed this position. In R v Clear7, the Court of Appeal held that “menaces” encompass threats of any action detrimental or unpleasant to the addressee and that it is immaterial whether the threatened act is lawful or unlawful, a reading that aligns with modern commentary emphasising the abusive deployment of pressure rather than the intrinsic lawfulness of the threatened act (Scaife 2013). Likewise, Lord Wright’s dictum in Thorne v Motor Trade Association8 confirmed that a threat to do an act lawful in itself, such as reporting a matter to the authorities, may amount to a menace when used as a vehicle for extortion (Lupton 2024).

Two clarifications follow for present purposes. First, the indifference of blackmail doctrine to veracity concerns the truth-value of the allegation or imputation that is threatened; it does not depend on any requirement that the threatened basis be factually correct. Secondly, deepfake sextortion commonly supplies the threatened imputation in the form of a fabricated sexual artefact that functions as putative evidence and is designed for circulation. That representational mechanism may intensify menace and may shape institutional assessment of threat-stage harm, even though the coercive logic remains, in principle, unaffected by truth-value. At the same time, blackmail does not provide a complete template for deepfake sextortion. Section 21 Theft Act 1968 requires a demand made with a view to gain or to cause loss, and many instances of deepfake sextortion are directed at outcomes that are not straightforwardly reducible to property gain or loss (for example, sexual compliance or enforced silence). The consequence is that the law’s recognition of veracity-indifferent coercion in blackmail does not resolve the separate question of how threats implemented through fabricated sexual material should be characterised and regulated where the blackmail elements are not clearly met.

When that settled framework is confronted with threats implemented by fabricated deepfake material, its explanatory power becomes uneven. A threat may coerce irrespective of veracity; the distinctive feature here is that coercion is implemented through a fabricated sexual artefact that functions as putative evidence and is designed for circulation. Three deficiencies become apparent. First, the law contains no express recognition of “threatening through fiction” as a mechanism of coercion in which the fabrication itself supplies the leverage, irrespective of whether any underlying conduct occurred (Edwards 2019; Allison 2024). Secondly, in the absence of a demand for property or money, doctrine provides no stable standard for identifying intent to cause distress through the threatened deployment of synthetic sexual content, as distinct from mere offensiveness or ill-judged expression; a broad conception of sexual content may avoid technical boundary disputes, but it does not, without more, supply a principled marker for coercive deployment (Citron and Franks 2014). Thirdly, although online abuse often facilitates dissemination, the coercive harm may crystallise before wide publication—through private threats, targeted transmission, or imminent upload—yet civil routes for anticipatory protection remain indirect and uneven. Defamation is structured around publication and serious reputational injury, with truth as a complete defence; it is therefore an awkward vehicle where the wrong is coercive leverage and the victim’s immediate need is prevention rather than post-publication vindication. Malicious falsehood similarly presupposes publication and typically pecuniary loss. Misuse of private information presupposes the disclosure of information properly characterised as private; a fabricated depiction may fall outside that frame where it does not disclose private facts about the claimant. Each cause of action therefore rests on predicates that deepfake coercion can displace (McGlynn et al. 2017).

These limitations justify recognising a distinct legal category: coercion through falsity. It captures the use of digitally fabricated yet credible sexualised representations as leverage to manipulate, intimidate, or extort. Its distinctive feature lies in the mechanism. Blackmail doctrine is concerned with unwarranted demands backed by menaces and is indifferent to the truth of the allegation or imputation that supplies the leverage; defamation is structured around publication of a falsehood. Coercion through falsity, by contrast, centres on a fabricated sexual artefact functioning as putative evidence and on the communicative use of that artefact to constrain the target’s options. The harm does not depend on the fabrication being widely believed, nor on any underlying “real” event, but on the victim’s rational apprehension that the depiction will be treated as credible and deployed against them (Laffier and Rehman 2023). The resulting subjugation is threat-stage and relational: autonomy is constrained by the prospect of reputational and social consequences that follow from plausible attribution.

This account does not criminalise “bad thoughts”, nor does it rest on controversial notions of preparatory liability. The actus reus is externalised conduct—creation, possession, or transmission of a fabricated sexual depiction combined with a communicated menace, conditional demand, or coercive threat—and the mens rea concerns the intention that the depiction be used as leverage. Proof will ordinarily turn on materials routinely relied on in digital offending: communications evidencing the menace or demand; the existence and provenance of the fabricated file; device and platform records; and expert analysis where authenticity, manipulation, or synthetic generation is disputed. A threat-stage offence in this domain is therefore not “on the road” to a more serious completed offence; it targets a completed wrong in itself, namely coercive constraint imposed through a fabricated sexual representation, whether or not publication subsequently occurs.

This form of coercion corresponds to the philosophical conception of coercion as the control of one agent’s choices by another through manipulation of the choice context (S. A. Anderson 2010). The deepfake perpetrator seeks not only pecuniary gain but the subordination of the victim’s will. The harm is therefore not exhausted by reputational damage. It includes the constraint of agency produced by credible threatened exposure, alongside predictable psychological, social and epistemic consequences that are imperfectly captured by existing doctrinal categories.

Judicial recognition of analogous harms is emerging, albeit unevenly. A threat to circulate intimate images can, of itself, generate anxiety and fear of a kind capable of amounting to substantial harm, even where no dissemination ultimately occurs (Campbell et al. 2022). In United States v Petrovic, Judge Berzon observed that threats can “enslave the will” by forcing a choice between intolerable alternatives9. In United States v Coss, Judge Moore accepted that a fabricated child-pornographic image may be as coercive as a real one10. The same organising idea is reflected in R v Jheeta11, where the Court of Appeal affirmed that coercion through a manufactured narrative can vitiate consent. In parallel, the jurisprudence of the European Court of Human Rights recognises that intimidation and psychological pressure may constitute an interference with autonomy and dignity12. Empirical evidence also indicates the scale of the problem: the International Organised Crime Threat Assessment 2024 reports a sharp increase in online sexual extortion targeting minors, and recent studies estimate that between 3.5 and 5 per cent of young people experience sextortion before adulthood, with girls disproportionately affected (Europol 2024b).

Deepfake fabrication therefore exploits credibility as a mechanism of control. By generating plausible falsity, it can constrain autonomy through the anticipatory force of threatened exposure. The harm may crystallise at the moment the threat is communicated, irrespective of the fabrication’s truth or its eventual publication. Coercion through falsity should accordingly be treated as a legally cognisable form of synthetic sexual harm (Pawelec 2022), warranting doctrinal recognition on a footing comparable to extortionary pressure, but calibrated to coercion implemented through fabricated sexual representations in contemporary digital communications.

2.5. Epistemic Injustice and the Denial of Harm

The emerging jurisprudence and practice on deepfake sextortion disclose an epistemic as well as a normative problem: how institutions assess credibility where the mechanism of coercion is synthetic and evidentially contestable. When complainants reporting synthetic sexual coercion are met with suspicion, evidentiary scepticism, or institutional indifference, the process risks inflicting what Fricker terms testimonial injustice: a wrong done to a person as a knower through credibility deficits rooted in prejudice and power (Fricker 2007). Testimonial injustice occurs when prejudice causes a hearer to assign deflated credibility to a speaker’s testimony, thereby wronging them specifically in their capacity as a knower (Holgado et al. 2021).

The point is not that institutional caution is illegitimate. It is that credibility assessment becomes normatively salient where doubt is distributed in patterned ways and operates as a gatekeeping device for protection, investigation, and redress. In that setting, disbelief can function as a continuation of coercive constraint rather than its aftermath. It compounds the harm by denying the legitimacy of the complainant’s account (Fricker 2007). For deepfake victims, this epistemic harm manifests through victim-blaming, normalisation, casual dismissal, and disbelief (Birrer and Just 2024).

This should not be treated as an argument for criminalisation. The claim here is narrower. Where deepfake material is deployed as an instrument of coercion, disbelief can impede access to existing criminal, civil, and regulatory routes and thereby entrench interference with autonomy. Epistemic injustice is therefore used as a diagnostic of institutional competence and error, not as a free-standing mandate to create new offences.

Courts and commentators have, in other contexts, recognised that credibility dynamics can extend or intensify the original wrong. Lord Steyn’s characterisation in R v A of post-assault disbelief as “a second rape”13 captures the proposition that epistemic invalidation can perpetuate harm through institutional practice. Hallett LJ’s dissent in R v Evans, highlighting the courtroom demand to “prove you did not enjoy it,” similarly treats adversarial scrutiny as a mechanism of re-traumatisation14. Related concerns about credibility, stereotyping, and social reception are also visible in People v Turner15 and State v Clevenger, which have been invoked to illustrate how the burden of contesting an imposed narrative can itself function as a secondary injury and as an obstacle to justice16. The relevance for deepfake sextortion is structural rather than factual: the complainant’s account is assessed within frameworks that may mis-specify what must be proved and that may allocate credibility deficits in predictable ways.

Synthetic sexual abuse is particularly susceptible to dismissal on the assumption that, absent physical contact or an authentic underlying act, the harm is illusory. Judicial and institutional responses framed as “it is only pixels”17 or “no real body was touched”18 exemplify this attitude in reported practice and triage, where fabrication is sometimes treated as a reason to downgrade seriousness or to deny an actionable wrong. These formulations are not treated here as doctrinal propositions; they indicate an evidential and classificatory tendency to privilege physical contact or authenticity when assessing harm and institutional competence. When such reasoning functions as a basis for institutional inaction or credibility discounting, it constitutes an epistemic wrong: disbelief becomes a mechanism through which primary coercion is converted into secondary, institutionally mediated injury.

The struggle for recognition of deepfake harms is playing out across comparative jurisdictions. In X v OnlyFans and B v Snapchat, the courts wrestled with the ontological status of “synthetic” injuries. Nicklin J’s obiter observation in X v OnlyFans that “the court cannot police phantoms”19 illustrates how evidentiary frameworks anchored in materialist assumptions obstruct redress for digital harms. Morgan J’s dissent in B v Snapchat, insisting that “the burden to authenticate non-existence is impossible,”20 crystallises the inversion of proof obligations borne by deepfake victims. These judgments exemplify what Fricker (2007) conceptualises as epistemic injustice: a structural inability of victims to convey harm within a system that presupposes their unreliability (See also Dotson 2011).

Empirical work reinforces this pattern of disbelief. Recent studies report that a substantial majority of deepfake victims—almost 87 per cent in one survey—are met with responses such as “it is not real harm” or equivalent dismissals of their experience as merely virtual (Ching et al. 2025). Such findings indicate that epistemic justice cannot remain a purely conceptual ideal; it must be operationalised as a structural virtue of institutions (Medina 2013). The persistent erasure of the credibility of synthetic victims signals not epistemic uncertainty but institutional moral failure, reflecting entrenched asymmetries in whose testimony is regarded as trustworthy and whose is discounted (E. Anderson 2012).

The point is particularly sensitive where targets are politicians. Democratic culture tolerates, and sometimes protects, aggressive and offensive commentary about public figures, including sexualised depictions. Recent UK institutional work on abuse and intimidation of MPs and candidates, and regulatory research on women in politics, emphasise that significant parts of the response space lie outside new criminal offences—through security practices, party duty-of-care measures, platform governance, and public education—precisely because overbroad criminal categories risk chilling protected speech (Speaker’s Conference 2025). The Scottish debate on criminalising misogyny, and the opposition it generated on breadth and vagueness grounds, underscores the need for tight offence definition where criminal law is invoked at all (Working Group on Misogyny and Criminal Justice 2022). The argument here therefore proceeds on a limiting basis: deepfake sextortion is analytically distinct from sexual lampooning or offensive political satire because it is constituted by coercive deployment—threat, conditional demand, or intimidation—rather than contribution to public debate.

A further dimension concerns the secondary injury caused specifically by epistemic injustice. Citron and Franks characterise deepfakes as “the second rape”, emphasising that disbelief and minimisation function not as a coda but as compounding harm. Fricker’s (2007) conceptual architecture can be extended to encompass “algorithmic testimonial injustice”, where credibility deficits are not only human but embedded through technical systems that filter, rank, and suppress reports of harm (Medina 2013). Medina’s account of “resistant silences” is instructive here: epistemic injustice operates not only through overt disbelief but also through institutional refusals to listen and through algorithmic exclusion from the “economies of credibility” on which recognition depends (Medina 2013).

A critical aspect of this landscape is the burden of proof for authenticity, which can operate as an epistemic trap. When individuals are compelled to prove they did not engage in acts that never occurred, they confront what Rini and Cohen term “illocutionary harm”—the coerced engagement in speech acts one would prefer to avoid (Rini and Cohen 2022). This forced denial subverts communicative agency, transforming complainants into perpetual defendants of their own non-existent conduct. The legal implications are practical: reporting is hindered by insufficient or complex tools and by authorities that discourage action because perpetrators cannot be identified (Flynn et al. 2022). The phenomenon also reflects what Fricker (2007) characterises as hermeneutical injustice—“[t]he injustice of having some significant area of one’s social experience obscured from collective understanding owing to hermeneutical marginalisation”—a dynamic further explored in contemporary analyses of image-based abuse (A. McGlynn 2023).

Requiring a deepfake victim to demonstrate that they are not who the image depicts converts authenticity into a juridical trap: the victim is asked to prove that they did not do what never occurred, and the inevitable difficulty of that task is then turned back against them as a further ground for doubt (Fricker 2007; Medina 2013). Insisting that a deepfake victim prove the falsity of an image replicates the ethical failure identified by Fricker: an epistemic wrong enacted through the denial of credibility (Fricker 2007). In the context of synthetic sexual abuse, disbelief cannot plausibly be treated as mere procedural caution where it functions as a systematic barrier to recognition and response. The necessary jurisprudential shift is therefore from treating “synthetic” status as a reason to downgrade seriousness to treating the epistemic and procedural position of the complainant as part of the harm profile. Only through relocating the focus from the ontological status of the image to the epistemic status of the disbelieved subject can law begin to address the full extent of harm in cases of deepfake sextortion.

3. The Cases of Deepfake: A Victim-Centred Analysis

Law may be conceived, in part, as an institutional practice of recognising legally salient interests and wrongs, and of translating that recognition into enforceable duties, remedial routes, and evidential allocations. On that understanding, the difficulty posed by deepfake sextortion is not simply that victims experience distress, but that existing categories may misclassify the wrong and thereby misallocate responsibility and proof. The injury cannot be reduced to reputational damage or emotional upset in any narrow sense. It includes interference with autonomy through coercive leverage, alongside epistemic consequences that shape whether the complainant is treated as credible, whether protective intervention is offered, and whether redress is realistically attainable. The harms at stake therefore include both the immediate constraint imposed by a single episode of synthetic coercion and the longer-term effects of forced sexual attribution on social standing, professional opportunity, and the victim’s capacity to control how her identity is represented and interpreted in digitally mediated settings.

This section develops a victim-centred analysis of deepfake sextortion, taking the high-profile case of Cara Hunter as an exemplar of the institutional problems addressed in this article. The case is used to show how existing legal and institutional frameworks may fail to register, and therefore to remedy, the distinctive harms generated by deepfake sextortion, and how those failures can themselves generate further injury. The analysis introduces the concept of synthetic re-victimisation to capture this dynamic: the repetition and amplification of harm through representational, institutional, and social practices that treat the victim not as a rights-bearing subject but as an object of suspicion, spectacle, or administrative deflection.

3.1. Cara Hunter: Deepfake as Public Shaming Technology

Cara Hunter, a Member of the Legislative Assembly in Northern Ireland, was subjected to a pornographic deepfake that rapidly circulated online. The synthetic video, which depicted her engaging in sexual activity, was fabricated with high fidelity using AI-driven facial replacement tools. The content was entirely falsified, yet was immediately received as truth—its credibility enhanced by Hunter’s status as a public figure and by the absence of meaningful verification mechanisms on the platforms through which it spread. The wrong did not lie in the authenticity of the material but in its plausibility; she was harmed not because the video was real, but because it was believable.

The deepfake targeting Hunter in April 2022 appears to have been generated using contemporary deep learning architectures, including generative adversarial networks (GANs) and autoencoders, which memorise and reproduce facial features and bodily movement with striking verisimilitude (Gambín et al. 2024). These systems operate iteratively: generators produce synthetic images based on encoded features, while discriminators refine those outputs by detecting artefacts and optimising towards ever more convincing fabrications (Sharma et al. 2024). High-fidelity facial swapping enabled Hunter’s likeness to be seamlessly grafted onto pornographic material, such that, to an untrained observer, the resulting footage was indistinguishable from an authentic recording (Dhanyalakshmi et al. 2025).

Dissemination occurred primarily through WhatsApp and social media, exploiting the frictionless virality and low-friction forwarding that characterise contemporary communication platforms21. Unlike earlier forms of image-based abuse, which typically retained some tether to underlying events or recordings, Hunter’s deepfake was wholly synthetic. Its persuasive force derived from the “credibility” conferred by machine learning techniques and from the gendered expectations attached to women in public life, rather than from any factual substratum. Monetisation was not overtly evidenced in this instance, although the attack aligns with a broader ecosystem in which bespoke deepfakes are commodified through illicit marketplaces; in Hunter’s case, the dominant motive appears to have been intimidation rather than profit.

Hunter’s immediate response was public and unequivocal. She repudiated the fabrication, engaged with media outlets and called for legislative reform. However, the institutional apparatus proved strikingly inadequate. The Police Service of Northern Ireland opened an investigation under provisions of the Justice Act (Northern Ireland) 2016, which criminalises the non-consensual disclosure of private sexual images, alongside the Protection from Harassment (Northern Ireland) Order 1997. Three years on, the perpetrator remains unidentified, with the investigation hindered by digital anonymity, cross-border data flows and the forensic challenges presented by sophisticated synthetic media.

In parallel, social media platforms exhibited the familiar pattern of delayed and partial response. Reliance on reactive, complaint-led content moderation policies produced slow and incomplete removal of the material. This inertia persisted despite the emerging duties of care crystallised in the UK’s Online Safety Act 2023, and it was exacerbated by the absence, at the time, of any specific offence in Northern Ireland law targeting the creation of sexually explicit deepfakes. Only with a 2025 Department of Justice consultation were proposals formally tabled to criminalise the non-consensual production of sexual deepfakes. The lag between technological capability and legal recognition is thus inscribed in the chronology of Hunter’s case.

According to Hunter, she was targeted with digitally forged sexual content as a deliberate act of intimidation in advance of electoral contests, with profound effects on her welfare and sense of personal security. The chilling effect of such sexual digital forgeries operates through silencing and isolation. It signals to women and girls in public life that visibility carries the latent threat of synthetic sexual humiliation, thereby deterring entry into, and continuance within, political and public spheres. The impact is twofold. At the individual level, women experience a contraction of their “space for action”, forced to navigate public life under the constant possibility of representational violation. At the structural level, this contraction reinforces male dominance in online spaces and public discourse, thereby depriving society of equal and diverse participation.

Öhman’s characterisation of deepfake sexual abuse as a coordinated system of actions rather than as isolated instances of fantasy is instructive here (Öhman 2020). On this view, the significance of any single video lies less in its impact on the individual depicted than in its contribution to a broader system that systematically degrades women and normalises their sexualised exposure as entertainment or sanction. Hunter’s experience, therefore, cannot be understood solely as a discrete attack on one politician; it should be read as part of a broader pattern in which synthetic sexual images function as tools of public shaming and gendered discipline.

Public reactions to the Hunter deepfake oscillated between disbelief, voyeurism and victim-blaming. Such responses reproduce familiar dynamics from the literature on image-based sexual abuse against women in politics, where victims are scrutinised, pathologised or mocked rather than believed (Sensity AI 2024). The Cara Hunter case thus lays bare the interface between technologically mediated falsity and its tangible social, psychological and reputational consequences. It exposes the failure of existing institutions to address what might be termed the “representational threat” of deepfakes: the capacity of synthetic sexual images to reconfigure a victim’s public persona and constrain her agency without any underlying physical act. Far from being an aberration, Hunter’s ordeal is paradigmatic. It underscores the need for statutory reform that directly addresses synthetic sexual abuse, for platform-level accountability commensurate with the scale and speed of dissemination, and for societal recognition of deepfake sextortion as a violation of fundamental rights rather than a regrettable by-product of digital culture. An equally deliberate evolution must match the evolution of deepfake technology in terms of legal, institutional, and social safeguards.

3.2. The Structure of Synthetic Re-Victimisation

Hunter’s experience exemplifies what this article terms synthetic re-victimisation. The phenomenon does not consist merely in the recirculation of a fixed set of abusive materials, as in classic “revenge pornography”, but in the capacity of generative systems to produce an indefinite sequence of novel, hyper-realistic fabrications from the same underlying biometric template. Each new iteration reactivates and intensifies the original trauma, generating a cycle of renewed exposure, disbelief and psychological harm (Worsley and Carter 2022). Emerging scholarship has begun to describe this as the repeated use of an individual’s likeness to construct new, fabricated representations, each capable of going viral and producing new waves of damage (Stavola and Choi 2023). Rather than a discrete incident, the victim confronts an ongoing process in which their digitised image functions as a renewable resource for abuse. The result is a state of chronic hypervigilance in which the individual must constantly anticipate the next assault on their digital self, while legal and regulatory responses remain confined to episodic interventions such as individual takedowns (Flynn et al. 2025).

The conceptual roots of synthetic re-victimisation lie in victimology and the literature on image-based sexual abuse, now increasingly extended to AI-mediated practices. Deepfakes can be situated within the broader category of non-consensual intimate imagery, overwhelmingly directed at women and typically sexualised in content (Ruvalcaba and Eaton 2020). In domestic abuse contexts, perpetrators use deepfakes as instruments of coercive control, superimposing a victim’s face onto explicit material and threatening dissemination in order to maintain psychological domination long after physical separation (Lucas 2022). This practice positions deepfakes as an emergent modality of gender-based violence, in which technologically enhanced credibility is mobilised to undermine autonomy and public trust in the victim’s denials (Akter and Ahmed 2025). Citron’s account of “cyber civil rights” underscores how such practices convert private violation into public spectacle and reinforce patterns of digital misogyny, with victims frequently deterred from seeking redress because they anticipate scepticism or ridicule (Citron 2009). Synthetic re-victimisation thus operates at the intersection of gendered power, technological affordances and institutional inaction.

Empirical work confirms both the scale and the cumulative nature of this harm. A 2023 United Nations report on technology-facilitated gender-based violence attributes approximately 90 per cent of non-consensual synthetic pornography to deepfake production, and links such material to offline consequences including stalking, employment loss and social exclusion, noting also that victims frequently under-report precisely because they expect not to be believed (United Nations Women 2023). In educational settings, a recent study on AI-driven bullying indicates that 13 per cent of school leaders have encountered deepfake incidents, predominantly targeting girls, with reported outcomes including heightened anxiety, withdrawal from peer networks and deterioration in academic engagement (Jackson et al. 2025). These data support the view that synthetic re-victimisation is not an abstract or speculative risk but a measurable pattern of repeated injury over time.

Existing legal frameworks struggle to accommodate this temporality. US jurisprudence in Ashcroft v. Free Speech Coalition22 invalidated prohibitions on “virtual” child pornography on the basis that purely synthetic images do not, in themselves, document real abuse. Congress responded through the PROTECT Act 2003,23 including the creation of offences directed at obscene visual representations and amendments to the statutory definition of child pornography to cover digital or computer-generated images that are, or are “indistinguishable from”, depictions of minors engaging in sexually explicit conduct.24 In United States v Williams, the Supreme Court upheld the Act’s pandering/solicitation approach25, which targets offers or requests made in a manner intended to cause the listener to believe that the material depicts real children. These developments, however, do not resolve the difficulties raised by adult deepfakes: the post-Ashcroft settlement is closely tied to child-protection rationales, obscenity thresholds, and speaker-belief/intent structures that do not map cleanly onto threat-stage coercion and anticipatory harm in sextortion contexts. This is why Ashcroft’s reasoning has been criticised as indirectly insulating adult deepfakes from regulation where the wrong lies in coercive leverage rather than the documentation of an underlying act (Pascale 2023). Justice O’Connor’s partial dissent already anticipated that morphed or fabricated images might generate “significant non-obscene harms” by eroding the distinction between real and fictional sexual content (Atisha 2025). Current reform proposals in Europe and elsewhere tend to rely on adaptations of existing tools—extensions of privacy or publicity rights, or torts framed as “synthetic defamation”—and on post hoc remedies such as takedown orders and delisting. Commentators warn that such approaches, focused on individual artefacts, address only the symptoms of an iterative and scalable pattern of fabrication and circulation, leaving intact the enabling conditions for repeated abuse (Ajder et al. 2019; Vaccari and Chadwick 2020).

A victim-centred account of synthetic re-victimisation therefore requires a shift from artefact-based liability rules to a framework that recognises the perpetuity and renewability of harm. Hunter’s case illustrates the point with particular clarity: removal of specific instances of the deepfake did not restore control over her image, because the underlying capacity to generate further fabrications remained untouched. Legal doctrines that treat each upload as an isolated event mischaracterise the wrong and underestimate its cumulative impact on autonomy, participation, and security. Synthetic re-victimisation, understood as an ongoing process rather than a single occurrence, calls for legal standards that address the continuing risk of re-fabrication and recirculation, and that place positive obligations on platforms and regulators to disrupt the conditions that sustain such cycles. Without such a reconceptualisation, victims remain effectively trapped in a form of digital purgatory, compelled to relive a technologically mediated harm that existing law is not yet structured to see, let alone to remedy.

3.3. Victim Disbelief and Institutional Silence

Victims who present synthetic sexual images as evidence routinely encounter epistemic resistance; the material is dismissed as inconsequential precisely because it is fabricated. However, credible deepfakes can inflict trauma comparable to that caused by “real” violations, eroding dignity and autonomy without any physical contact (Europol 2024a). The insistence on verifiable truth as a precondition for legal recognition falters in this context, since the relevant harm does not stem from factual accuracy but from perceptual credibility. This mismatch sustains what Sweet terms institutional gaslighting, in which victims are required to authenticate harms whose very purpose is to destabilise reality, thereby extending and entrenching the abuse (Sweet 2019).

Law’s evidentiary realism, particularly in doctrines developed for a pre-digital environment, demands a demonstrable link to material reality. In image-based sexual abuse, this has translated into a requirement of authenticity. Defamation and privacy frameworks in many European jurisdictions were drafted on the implicit assumption that the impugned material records an underlying event. A German court operating under the older provisions on “violation of intimate privacy”, for example, may struggle to subsume a provably computer-generated image under that heading, pushing the complainant towards ill-fitting causes of action such as defamation, which focus on reputation rather than sexual and psychological autonomy (German Criminal Code 1998). The result is an authenticity trap: the victim must first prove that the image is fake in order to gain access to any legal remedy, a process that is technically complex, costly, and frequently re-traumatising. That demand mirrors the abuser’s tactic, which is precisely to construct a reality so plausible that it destabilises the victim’s epistemic footing. McGlynn rightly observes that a focus on the creator’s intent or the image’s provenance occludes the harm experienced at the point of receipt; the core wrong lies in the non-consensual dissemination of a sexualised representation of the self, irrespective of whether the depiction corresponds to any underlying act (P. C. McGlynn 2022).

European regulatory developments both acknowledge and reproduce these tensions. The European Union’s (2024) Directive on Combating Violence Against Women and Domestic Violence expressly classifies non-consensual deepfakes as “image-based violence” and requires Member States to criminalise their creation and distribution, coupled with remedies such as content removal (Directive (EU) 2024/1385). Implementation, however, remains uneven. Denmark’s 2025 Deepfake Bill, for instance, proposes privacy-based torts but conditions relief on evidentiary authenticity, effectively mandating forensic validation of synthetic material (Willige 2025). In the Netherlands, analogous proposals ground claims in reputation law, yet victims report institutional dismissal where images evade detection tools or do not fit within pre-existing offence categories (Groothius 2021). Hunter’s case sits squarely within this area. Her sustained advocacy for legislative reform underscores how the absence of doctrinal recognition relegates victims to the position of supplicants, required to argue for harms that the legal framework does not yet conceptualise, while fabricated evidence simultaneously undermines their credibility in related investigations (Europol 2024a). A comparative study of EU Member States confirms the fragmentation: victims in jurisdictions such as Poland and Hungary are often confined to civil defamation claims that prioritise demonstrable “real” harm over the psychological and epistemic injuries characteristic of synthetic sexual abuse (Mania 2022). The pattern that emerges is one in which disbelief and institutional silence are not incidental but constitutive; they form part of the structure of synthetic re-victimisation that deepfake sextortion both exploits and amplifies.

3.4. Legal Silence as Structural Harm

The legal silence surrounding deepfake sextortion operates as a form of structural harm rather than a simple gap in doctrinal coverage. The absence of clear statutory provisions addressing the non-consensual creation and use of synthetic sexual imagery leaves victims without a stable foundation for redress. Fragmentation across criminal law, civil claims, data protection regimes and platform policies further produces a regulatory labyrinth that most victims are ill-equipped to navigate. Hunter’s attempts to secure accountability encountered not overt bad faith, but a diffuse structural incapacity: police officers lacked appropriate investigative tools, platforms lacked meaningful legal incentives to act, and legislators lacked a sense of urgency. This constellation of omissions constitutes what may be termed regulatory invisibilisation: the systemic rendering of synthetic sexual harm as lying outside the purview of actionable law, despite its tangible social and psychological impact.

Zemiological and victimological scholarship provide a vocabulary for understanding this silence as social harm embedded in institutional design. Macdonald and Peacock argue that harms facilitated or tolerated through legal and policy arrangements amount to “social harms” that normalise the invisibility of gendered violence (Macdonald and Peacock 2025). Feminist strands of victimology similarly locate image-based abuse within patriarchal structures that render women’s digital bodies disposable, and that amplify coercion through falsity by privileging perceptual plausibility over empirical truth (Ali et al. 2025; de Silva de Alwis 2024). Walklate’s work on crimes against women emphasises that structural harms often arise not only from discrete acts, but from institutional failures and the subjective terror of violation, and that victims consequently endure “epistemic injustice” when their testimonies are dismissed or downgraded (Walklate 2017). Deepfake sextortion fits squarely within this frame: legal institutions systematically under-recognise synthetic sexual harm, thereby reinforcing the very asymmetries that perpetrators exploit.

Hunter’s experience in Northern Ireland makes these dynamics concrete. After a deepfake video superimposing her face onto explicit content was created and accompanied by extortionate threats, she engaged with law enforcement under a framework that had been designed for non-synthetic intimate images. The Justice Act (Northern Ireland) 2016 did not clearly criminalise the creation of fabricated sexual imagery absent disclosure, and investigative authorities cited evidentiary gaps rather than offering a coordinated response. Hunter’s subsequent advocacy revealed a pattern of structural incapacity: no dedicated institutional mechanism existed to address synthetic sexual coercion, and responsibility was diffused across agencies and regulatory domains. Structural strain theory suggests that such regulatory voids foster anomie, allowing emergent forms of gendered cybercrime to flourish in the space between social expectations of protection and the law’s actual reach (Agnew 1992). Deepfake sextortion thus operates in a zone where offenders rationally anticipate impunity and victims rationally anticipate institutional inertia.

Normative proposals in recent scholarship converge on the need to move beyond evidentiary realism toward standards that foreground experiential harm. Dushi, Berdufi and Karagianni argue for a shift towards legal tests framed in terms of “perceived harm” and “violation of sexual integrity”, with harm principles integrated directly into statutory language so that technologically mediated abuses do not fall through doctrinal cracks (Dushi et al. 2025). Such an approach would reorient legal analysis around the phenomenology of harm rather than the verifiability of images. Hunter’s case demonstrates that what is decisive is not whether a deepfake is empirically “real”, but whether its social meaning is mobilised to violate autonomy, constrain agency and erode dignity. Legal silence in the face of such practices entrenches structural coercion through falsity, as institutional inaction and doctrinal fragmentation combine to validate the message that synthetic sexual abuse is, at best, marginal. Only a framework that centres victims’ experiential realities and treats synthetic sexual harm as within the protected core of sexual and communicative self-determination can begin to dismantle this structural harm and meet the challenge posed by coercion through falsity in the digital age.

4. Doctrinal and Regulatory Analysis: Interlocking Frameworks in England, Wales and Northern Ireland

Deepfake sextortion presents a systemic challenge for contemporary legal regulation. It sits at the intersection of multiple bodies of law—sexual offences, blackmail and cybercrime, communications and harassment offences, intimate-image abuse, and online platform regulation—yet it is not straightforwardly accommodated by any of them. The resulting injury is hybrid: reputational and sexualised, psychological and economic, mediated through digital artefacts. Because the wrong is constituted by credible fabrication rather than factual exposure, its juridical features cut across the organising assumptions on which existing rules are built. No single doctrinal regime, considered in isolation, therefore yields a coherent or comprehensive response.

This section undertakes a critical analysis of the principal frameworks in England, Wales and Northern Ireland, with the Online Safety Act 2023 as the central regulatory anchor. It examines, first, the extent to which criminal offences concerned with threats, extortion and intimate-image abuse capture coercion through falsity; secondly, the capacity of communications, harassment and platform-governance regimes to address synthetic sexual harm; and, thirdly, the availability of civil and regulatory remedies capable of recognising anticipatory, psychological and epistemic injury. The analysis shows that, both individually and in combination, these regimes contain conceptual and operational gaps that leave victims of deepfake sextortion without clear routes to redress, stable recognition of the wrong, or effective protection against recurrence.

England, Wales and Northern Ireland: Fragmented Frameworks and Doctrinal Limits

This section examines how deepfake sextortion is positioned within the United Kingdom’s evolving legal framework, with the Online Safety Act 2023 (“OSA”) as the central regulatory framework. The OSA was indicated as a systemic response to online harms, aligning criminal law, platform regulation and user protection. However, when examined against the specific phenomenon of deepfake sextortion, the framework reveals a series of conceptual and structural misfits. The statutory scheme remains organised around discrete categories of “illegal content” and platform “systems and processes”. In contrast, the harm analysed earlier in this article is continuous, relational and cumulative: a form of coercion through falsity that destabilises autonomy and corrodes the conditions under which victims can participate in social and political life.

(a) 

The Online Safety Act 2023 as Systemic Anchor

OSA imposes safety duties on regulated user-to-user and search services in respect of “illegal content”, with enhanced requirements for priority offences and pornography services. Image-based sexual abuse falls within the Act’s illegal-harms architecture insofar as relevant offences are engaged, and the regime is designed to make providers operationally responsible for preventing and responding to illegal harms online. Ofcom frames the regime in these terms: online providers are now legally required to protect users from illegal harm, principally through risk assessment, proportionate mitigation, reporting, and enforcement mechanisms rather than by creating an autonomous body of substantive online wrongs (Ofcom 2024).

It is not whether the OSA “addresses deepfakes” in the abstract, but how far an offence-keyed, service-centric risk-management regime can track the distinctive temporality and mechanism of deepfake sextortion: coercion implemented through fabricated sexual artefacts as putative evidence, often producing harm at the threat stage and often involving toolchains and specialist services outside mainstream platform governance. Kira’s analysis of non-consensual intimate deepfakes (‘NCID’) is relevant here. The OSA is framed around content on regulated services and the identification of an underlying criminal offence; it does not directly regulate the act of creating NCID, nor the AI tools and upstream services that make fabrication cheap and scalable (Kira 2024). Ofcom’s guidance can sharpen providers’ risk assessment and improve decision-making about illegal content, but it cannot extend the statute’s locus of responsibility beyond the regulated service perimeter, nor can it remove dependency on the coherence and coverage of the underlying criminal law categories that supply the “illegal content” trigger (Ofcom 2024).

This limitation matters when the OSA is read alongside the Law Commission’s reclassification of “intimate image abuse” as a form of sexual violation rather than a subset of communications or public order offending (Law Commission 2022). The Commission’s analysis centres the wrong in the non-consensual appropriation and sexualisation of a person’s image, irrespective of whether the material is real or fabricated. The OSA can reflect that shift only indirectly, through the way priority offences are listed and operationalised; it does not itself supply a general organising principle that treats image-based abuse as a modality of sexual violation across the regime. The result is that “recognition” of the wrong remains mediated through offence lists and content-judgement exercises, with predictable pressure points where deepfake sextortion is experienced primarily as threatened exposure, identity capture, and anticipatory constraint rather than as a post-publication moderation problem.

McGlynn, Woods and Antoniou make a related point in their critique of the OSA’s pornography provisions. Their argument is that the legislation embeds a narrow and fragmented conception of sexual harm, combining obscenity-inflected concerns with child protection while marginalising the structural character of violence against women and girls (McGlynn et al. 2024). The same analytical concern translates to deepfake sextortion insofar as regulatory attention is channelled towards identifiable items of content and compliance processes, while the survivor’s primary need may be prevention at the point of threat, rapid interruption of coercive leverage, and durable remedies that do not depend on repeated identification and removal of iterated content.

The Children’s Commissioner’s work on nudification tools and sexually explicit deepfakes reinforces the point from a child-protection perspective. Evidence submitted to the Commissioner indicates that AI-powered “nudification” and deepfake services are readily accessible and can be used to create synthetic sexual images at scale; downstream moderation on mainstream platforms does not, without more, constrain the upstream supply of tools and specialist sites that drive production (Children’s Commissioner for England 2025). This supports the case for “safety by design” obligations directed at development and deployment pathways, rather than reliance on downstream content governance alone.

Read together, these analyses indicate a recurring pattern. The OSA—amplified by extensive Ofcom guidance—implements a risk-based compliance model intended to apply and enforce existing illegal-harm standards more effectively. The residual question is whether that model, even when functioning well, can capture deepfake sextortion as a process of coercion through falsity and synthetic re-victimisation: harm crystallising at the threat stage; leverage created by fabricated evidential artefacts; and recurrence enabled by tool-driven replication. On this account, the critique is not that the OSA lacks regulatory machinery, but that its offence-triggered, service-centred and content-management orientation leaves identifiable limits in addressing threat-stage coercion and upstream enablement without further reform.

(b) 

Sexual Offences Act 2003, Intimate Image Offences and the Crime and Policing Bill

Amendments made through and alongside the OSA have significantly reworked the Sexual Offences Act 2003 (“SOA 2003”), particularly through the creation of new intimate-image offences. In England and Wales, the new structure includes: (i) a strict-liability offence of sharing an intimate photograph or film without consent, with a defence of reasonable excuse; and (ii) aggravated offences where sharing or threatening to share intimate images is carried out with intent to cause distress, alarm or humiliation, or for purposes of sexual gratification. Deepfake images purporting to depict a real, identifiable person fall within these offences, and threats to share such images now clearly constitute criminal conduct (Law Commission 2022).

This reform undoubtedly marks progress. Victims no longer need to prove the perpetrator’s intent to cause distress to establish a basic offence, and deepfakes are expressly recognised as a form of intimate image abuse. Nevertheless, as Kira notes, the structure of liability remains tethered to dissemination and threat: creation of sexually explicit deepfakes, without subsequent sharing or threat, is not generally criminalised in England and Wales, even where images are plainly produced for extortion or later use (Kira 2024). The UK Government now presents the Crime and Policing Bill as completing this framework, implementing the Law Commission’s recommendations on the taking and recording of intimate images, and filling perceived gaps left by the OSA-driven reforms (Ministry of Justice 2025).

The Bill repeals the existing voyeurism provisions in sections 67(3) and 67A(2) SOA 2003 and replaces them with three new offences: a “base” offence of taking or recording an intimate photograph or film without consent or reasonable belief in consent; an offence of taking or recording without consent with intent to cause alarm, distress or humiliation; and an offence of taking or recording without consent and for the purpose of sexual gratification of oneself or another (Ministry of Justice 2025). It also introduces equipment offences, criminalising the installation, adaptation, preparation or maintenance of equipment with the intention of enabling any of these taking/recording offences. Doctrinally, this marks an important shift away from the narrow voyeurism paradigm: non-consensual imaging is recognised as a wrong, and the law accepts that harm begins at the point of capture or enabling, not only at the point of later misuse.

However, once deepfake sextortion is placed at the centre of analysis, the limits of this settlement become apparent. The new offences are drafted in terms of “taking or recording an intimate photograph or film”, language historically associated with capturing real-world scenes. The government’s own definition of intimate image abuse covers images that “show or appear to show” a person in an intimate state, which would allow courts to interpret the act of generating a deepfake as “recording” an intimate image (Ministry of Justice 2025). That interpretation is defensible and consistent with the Law Commission’s analysis. However, the absence of any express reference to synthetic media raises concerns about legality and foreseeability, particularly in a context where criminalisation is said to rest on a “holistic package” of offences.

A further structural limitation lies in the Bill’s equipment offences. Their paradigm case is the installation of hidden cameras in bathrooms or changing rooms. Generative AI tools, “nudification” services and web-based deepfake engines do not fit comfortably within this conception of “equipment”. The legislative focus remains on the physical means of capturing reality, rather than the digital infrastructure that fabricates plausible falsity. We argue that technology-facilitated abuse underscored precisely this point: without direct regulation of the tools and services that industrialise deepfake production, criminalisation of individual acts of sharing or taking risks remains largely symbolic. The Crime and Policing Bill does not yet confront that problem.

The claimed symmetry between the SOA 2003, as amended by the OSA, and the new taking/recording offences therefore warrants scrutiny. For conventional intimate images, the combination of sharing, threatening, taking, recording and equipment offences does produce a more coherent and vertically integrated regime. In the deepfake context, however, the regime still leaves the core act—the synthetic fabrication of a credible sexual image in the absence of any underlying event—only indirectly addressed. The Bill strengthens protection against certain forms of non-consensual imaging and aligns the law more closely with the Law Commission’s conception of intimate image abuse as a sexual autonomy offence. However, it does not resolve the more fundamental question identified earlier in this article: how the criminal law should treat coercion through falsity when the harm is anticipatory, epistemic, and relational.

The Northern Ireland context, as analysed by the Northern Ireland Human Rights Commission (“NIHRC”), illustrates further fragmentation. Department of Justice proposals would criminalise the intentional creation, sharing and threatening to share sexually explicit deepfakes of adults without consent, where the perpetrator acts with intent to cause humiliation, alarm or distress, or for sexual gratification (Northern Ireland Human Rights Commission 2025). The NIHRC welcomes the recognition of deepfake abuse as a human-rights issue engaging Articles 8, 10, 13 and 14 ECHR, CEDAW and the Istanbul Convention, yet raises two doctrinal concerns that resonate beyond Northern Ireland.

First, the reliance on specific motives risks under-criminalisation. Many instances of deepfake sextortion involve mixed motives or motivations that are difficult to evidence, such as control and coercion in an intimate relationship. The NIHRC therefore recommends the introduction of a summary-only base offence of intentionally creating, requesting, sharing or threatening to share a sexually explicit deepfake without consent or reasonable belief in consent, regardless of motive (Northern Ireland Human Rights Commission 2025). The proposal aligns closely with the logic of the strict-liability-based base offence adopted in England and Wales and addresses the structural problem that the most pernicious harms are frequently inflicted in circumstances where specific intentions are hard to prove.

Secondly, the NIHRC highlights the risk of intra-UK disparity. A model in which England and Wales provide a strict-liability base offence, while Northern Ireland relies solely on hybrid offences requiring proof of specified motives, would leave victims in Northern Ireland with a materially weaker level of protection, despite facing identical cross-border digital harms (Northern Ireland Human Rights Commission 2025). Such divergence raises questions under Articles 13 and 14 of the ECHR regarding effective remedies and non-discrimination in the enjoyment of Convention rights. For deepfake sextortion victims, the result is a patchwork of protection contingent on territorial boundaries that bear little relation to the transnational circulation of synthetic images.

(c) 

Communications and Harassment Offences: Residual and Misaligned

Legacy communications offences and harassment law continue to operate as residual routes for victims of deepfake sextortion. The Malicious Communications Act 1988 and relevant provisions of the Communications Act 2003 criminalise sending messages that are grossly offensive, indecent, or knowingly false, where the sender intends to cause distress or anxiety. The Protection from Harassment Act 1997 (“PHA 1997”) addresses “courses of conduct” that amount to harassment, and provides both criminal sanctions and a civil cause of action for victims seeking injunctions and damages.

These frameworks capture some instances of deepfake abuse, particularly where perpetrators directly send deepfake images or threats to victims, or repeatedly target them through multiple communications. However, their conceptual focus remains on messages rather than images as such, and on discrete episodes of “distress” or “alarm” rather than the ongoing, structural nature of synthetic sexual coercion. Requirements of a “course of conduct” under the PHA 1997, and of transmission and intent under the communications statutes, make these offences poorly suited to one-off but devastating sextortion threats, or to harms that arise predominantly from the viral circulation of content among third parties rather than direct communications with the victim (Equality Now and Alliance for Universal Digital Rights 2024; House of Commons Women and Equalities Committee 2025).

Critically, these statutes were drafted for a pre-AI environment. Their underlying assumptions are that harmful communication is discrete, traceable and attributable to an identifiable sender. Deepfake sextortion exploits precisely those features that lie outside this paradigm: anonymous or pseudonymous actors; decentralised and layered distribution; and the weaponisation of victims’ images in networks where the perpetrator may no longer need to communicate directly with the target. Reliance on these provisions, therefore, reinforces the dynamic identified earlier in the article: victims must fit their experience into legal categories that misdescribe both the harm and its mechanism, leading to institutional scepticism and a sense of legal unreality.

(d) 

The Online Safety Act Reconsidered: Structural Limits and Missed Opportunities

Positioning the OSA within this wider patchwork clarifies its strengths and limitations. It offers, in principle, a means of moving beyond purely reactive, victim-led enforcement under criminal and civil law, towards an ex ante regime of structural duties on platforms. However, the three sets of materials considered here—Kira’s doctrinal critique, McGlynn et al.’s pornography-focused analysis, the Children’s Commissioner’s investigation, and the NIHRC’s human-rights framing—all suggest that the law’s current design falls short of what a deepfake sextortion paradigm requires.

Kira’s central claim is that the OSA’s focus on platform “systems and processes” cannot compensate for the absence of robust, clearly defined offences governing NCID and the tools that produce them. Platforms are required to assess the risk that users might encounter illegal content, including intimate-image abuse, and to adopt proportionate mitigation measures. However, they are not required to prevent the availability of services whose primary function is to generate such content (Kira 2024). AI-powered deepfake and nudification tools remain largely outside the legislation’s direct reach, even though they constitute the technological infrastructure of deepfake sextortion.

McGlynn, Woods and Antoniou show that even within its core pornography provisions, the OSA retains a fragmented, content-centric logic. Part 5 places duties on pornography services to prevent children from encountering pornographic content, but does not require those services to prevent or redress non-consensual image-based abuse beyond the general illegal-content duties (McGlynn et al. 2024). The overall effect is an asymmetry in which the law expends significant regulatory energy on classifying and age-gating pornography, while leaving the structural conditions that enable image-based sexual abuse—including deepfake sextortion—comparatively under-regulated.

The Children’s Commissioner’s report underscores how this model fares in practice for children and young people. Evidence of widespread use of nudification and deepfake tools illustrates that generative AI services now operate as de facto pornography and abuse infrastructures (Children’s Commissioner for England 2025). However, the OSA’s current structure does not clearly mandate their restriction or removal, nor does it comprehensively integrate age assurance and safety-by-design obligations for AI tools themselves, as distinct from the platforms on which outputs may later circulate.

The NIHRC’s intervention adds a further layer of critique grounded in international human-rights law. Deepfake sextortion is treated as a form of digital gender-based violence that engages not only Articles 8, 10, 13 and 14 ECHR, but also obligations under CEDAW and the Istanbul Convention to adopt victim-centred, trauma-informed and gender-sensitive legislative responses (Northern Ireland Human Rights Commission 2025). A regime that relies predominantly on victims navigating fragmented criminal and platform complaint mechanisms, while leaving intact an underlying commercial ecosystem of deepfake services, struggles to satisfy the requirements of effectiveness, proportionality and non-discrimination in the enjoyment of Convention rights.

These strands of analysis point toward several doctrinal and policy refinements. First, intimate-image and deepfake offences need a clear, strict-liability base that covers intentional creation, request, sharing and threatening to share sexually explicit deepfakes without consent, without requiring proof of motive, accompanied by aggravated forms reflecting extortion, coercive control and sexual gratification. That structure aligns with the Law Commission’s reorientation of intimate image abuse as a sexual autonomy offence, and with the NIHRC’s concern that motive-based models leave harmful conduct unpunished where evidential thresholds cannot be met.

Secondly, the OSA’s safety duties require re-specification considering deepfake sextortion. A content-neutral duty on regulated services, including search engines, app stores and payment intermediaries, to prevent the provision and promotion of tools whose primary or predominant function is the creation of non-consensual sexually explicit deepfakes would give legal effect to the calls from the Children’s Commissioner and NIHRC for bans or severe restrictions on “nudification” platforms (Children’s Commissioner for England 2025; Northern Ireland Human Rights Commission 2025). Such an approach would also reflect the “business and human rights” framework, requiring companies to undertake due diligence across the lifecycle of AI systems that materially facilitate image-based sexual abuse.

Thirdly, Ofcom’s guidance and supervisory practice under the OSA should be framed so that providers’ risk assessments and mitigations are capable of identifying the coercive, threat-stage dynamics of deepfake sextortion rather than treating it solely as a post-publication content problem. The regulatory architecture is service-centred and procedural: it operates through risk assessment, systems design, reporting pathways, and enforcement of compliance obligations. The criminal justice response to perpetrators—investigation, charging decisions, and prosecution—necessarily falls elsewhere. Within that allocation of functions, the risk is that compliance metrics that privilege aggregate reductions in exposure to categories of illegal content will understate the intensity and persistence of harm borne by a comparatively small cohort of victims whose images are repeatedly weaponised, and will therefore misdirect mitigation priorities.

Finally, coherence across jurisdictions remains institutionally relevant for investigation and enforcement because deepfake sextortion is routinely cross-border in production, hosting, and targeting. Divergent offence definitions and procedural routes can impede cooperation and produce uneven practical protection. That concern does not presuppose that each jurisdiction should adopt broad, harm-based offences. The Scottish experience with proposals to criminalise misogyny after consultation illustrates the political and drafting constraints that attach to high-level criminal categories and the reluctance to legislate for contested harms at generality (Scottish Government 2025). The case advanced here is narrower: it concerns coercive deployment of fabricated sexual material as leverage. The emerging Northern Irish reforms, informed by NIHRC’s recommendations, illustrate both the potential of a rights-conscious approach and the risk that protection remains uneven unless a clearly defined base offence and associated enforcement routes are adopted.

Deepfake sextortion exposes the limits of a regulatory settlement that treats synthetic sexual abuse as a subset of “online safety” or “illegal content management”. The OSA provides an important systemic anchor, yet without substantive and institutional recalibration along the lines sketched above, the regime will continue to encode the very epistemic and structural misrecognition identified in earlier sections of this article. Victims will remain compelled to prove their harm within legal categories that struggle to see them.

5. Rethinking Remedies and Designing Accountability

Deepfake sextortion exposes a fundamental mismatch between the harms identified in the earlier sections of this article and the institutional tools currently available in the United Kingdom. The law remains organised around categories that presuppose physical contact, factual disclosure, or repeated conduct, whereas deepfake sextortion operates through plausible fabrications, singular threats, and technologically mediated asymmetries of power. The result is not merely under-enforcement, but a patterned failure of recognition: victims encounter legal structures that cannot “see” synthetic coercion as a primary wrong. This section argues that a credible response requires a shift from reactive, content-focused regulation to an integrated accountability model that treats synthetic sexual harm as a first-order concern of criminal law, platform governance, and remedial design.

5.1. Institutional Inertia and Conceptual Blindness

Earlier sections showed that deepfake sextortion produces harm at the point of threat, through the credible prospect of exposure and the enforced reorientation of the victim’s social and professional life. However, most UK legal routes still assume that injury crystallises only when material is disseminated or when “real” intimacy has been exposed. That assumption structures both criminal liability and access to civil remedies. The new intimate image offences in the Sexual Offences Act 2003, as amended, are triggered by sharing or threatening to share intimate images; communications offences turn on the sending of messages; harassment law relies on a “course of conduct”. A single, serious sextortion threat involving synthetic imagery may therefore fall awkwardly between categories or require victims to stitch together multiple causes of action to capture what they experience as a continuous, coercive practice.

This doctrinal architecture is underpinned by what the earlier conceptual framework described as evidentiary literalism: the tendency to treat authenticity and publication as preconditions for legal concern. Synthetic sexual images are often treated as less serious because they are “not real”, even though reputational collapse, occupational consequences, and psychological trauma depend on what third parties are likely to believe, not on whether an underlying sexual act occurred. Institutional scepticism towards synthetic harms reproduces the epistemic injustices already identified in relation to sexual violence more generally. Victims must not only explain the abuse but also persuade decision-makers that an event which “never happened” can nonetheless destroy relationships, careers, and a sense of safety.

This dynamic resonates with feminist critiques of the criminal law’s treatment of sexual violence, which have long highlighted the figure of the “perfect victim” and the law’s tendency to individualise structural harms. Syndromic or trauma evidence has historically been used to exceptionalise certain victims rather than to expose the ordinary operation of domination and coercion. In the deepfake context, disbelief takes a distinct form: institutions privilege the ontological status of the image over the social reality it generates, and in doing so, pathologise victims’ responses as overreactions to “mere” digital artefacts. The injury is compounded when police or prosecutors characterise cases as low priority, technically infeasible, or fundamentally civil in nature.

Technological capacity gaps exacerbate this conceptual myopia. Police forces and prosecutors’ offices are expected to respond to sophisticated generative systems with investigative tools and training designed for an earlier digital era. Deepfake detection and attribution remain unevenly distributed, reliant on specialist knowledge and expensive forensic methods, while many forces lack dedicated units or protocols for synthetic image abuse. Victims report being told that nothing can be done because the perpetrator has used anonymous accounts, virtual private networks, or encrypted messaging. The message, in effect, is that technologically complex harms fall outside the routine remit of law enforcement. That stance is not neutral. It functions as a de facto prioritisation decision in which synthetic sexual abuse is quietly displaced onto under-resourced victim support organisations or left entirely unmanaged.

The combined effect is institutional inertia: legal and investigative practices lag behind conceptual and technological developments, and the burden of navigating fragmentation falls on victims. The central question for reform, therefore, is not only how to adjust individual offences or procedures, but how to redesign the system so that synthetic sexual harm is recognised as a core concern rather than an awkward outlier.

5.2. Platform Governance Under the Online Safety Act

The Online Safety Act 2023 is often presented as the United Kingdom’s systemic answer to online harms. Section 4 showed that, for deepfake sextortion, the law functions as an incomplete anchor. Its duties of care on user-to-user and search services are structured around categories of “illegal content” and specified “priority offences”, with Ofcom empowered to issue codes of practice and enforce compliance. Image-based abuse offences, including those covering intimate deepfakes, are designated as priority offences; in principle, platforms must assess the risk of such content occurring, implement mitigation measures, and operate reporting and takedown systems.

We argue, however, that architecture remains heavily content- and exposure-focused. Duties are triggered by the risk that users will encounter illegal content on a service; enforcement is organised around reducing such encounters and improving the efficiency of removal processes. Deepfake sextortion, as analysed earlier, may generate its most acute harm before any content is uploaded or shared on a mainstream platform. Threats are often communicated through private messaging, semi-closed groups, or channels peripheral to the extensive, regulated services that form the core of the OSA regime, where a threat consists of the promise that content will be posted in the future—perhaps to a pornography site beyond UK jurisdiction—the immediate harm arises from the coercive pressure on the victim, not from the presence of content within a regulated service’s user base.

McGlynn et al.’s critique of the OSA’s pornography provisions is particularly salient here (McGlynn et al. 2024). Their analysis shows that the legislation conceptualises pornography primarily through child-protection and obscenity-inflected lenses, while failing to articulate a coherent account of pornography-related harms as forms of violence against women and girls (McGlynn et al. 2024). The fragmented conception is replicated in the treatment of intimate image abuse: deepfakes are absorbed into existing content categories without a broader recognition that synthetic sexual imagery can operate as a tool of gendered domination. The result is a regime that emphasises age assurance, classification, and access control, but does not adequately incorporate obligations to prevent or remediate non-consensual sexualisation and coercion.

The Children’s Commissioner’s work on “nudification” tools and sexually explicit deepfakes exposes a further structural limitation. Evidence submitted to the Commissioner demonstrates that children and young people can access AI tools that strip or sexualise images with minimal friction, often via app stores or search engines rather than through traditional social media platforms (Children’s Commissioner for England 2025). However, OSA’s focus remains on the latter. App stores, specialist deepfake generators and dedicated pornography sites sit at the edges of the regulatory perimeter, subject to general duties but not to a tailored regime that reflects their central role in the production and dissemination of synthetic sexual imagery.

Ofcom’s emerging guidance on deceptive deepfakes and attribution tools may, in time, provide a more granular framework for detection, labelling, and provenance (Ofcom 2025). At present, however, these developments operate within a discretionary space. Platforms retain broad latitude to decide which technical tools to adopt, how to calibrate detection thresholds, and how to balance responsive takedown with false positives and privacy concerns. This discretion is compounded in relation to private messaging and encrypted channels, where safety duties must be squared with commitments to confidentiality and freedom of expression.

The core difficulty is that the OSA treats platforms as managers of content risk rather than as gatekeepers of an infrastructure that enables certain forms of abuse to scale. That distinction matters. Deepfake sextortion is enabled not only by the presence of abusive content on a given service, but by the ready availability of generative tools, the design of recommendation and search systems, and the commercial incentives that sustain sites specialising in synthetic pornography. A regime that locates responsibility primarily in the speed and accuracy of takedown will inevitably lag behind harms that materialise at the point of threat, in semi-private spaces, or through off-platform circulation.

5.3. Technological Architecture of Evasion and the Limits of Enforcement

Deepfake sextortion exploits features of the contemporary digital environment that strain traditional enforcement methods. Threats are often issued from pseudonymous accounts, with perpetrators demanding payment in cryptocurrencies or other privacy-enhancing instruments. Synthetic content may be generated using open-source models or consumer applications, saved locally, and never uploaded to a platform within the reach of UK regulators. Where material is posted, it may be to services hosted offshore, with rapid mirroring and re-uploading that frustrate takedown efforts.

These features do not render enforcement impossible, but they do make clear that a strategy centred on identifying individual perpetrators and removing specific items of content will struggle to deliver meaningful deterrence. The investigative workload required to trace payments, attribute deepfake generation to particular devices, and secure cross-border cooperation is considerable. Resource constraints, competing priorities, and variable digital forensics capacity across police forces mean that only a small fraction of cases will receive that level of attention. Victims are acutely aware of this reality; it informs both their reluctance to report and their sense that the law does not take synthetic sexual abuse seriously.

The same technological affordances also complicate evidential practice. Synthetic images can be repeatedly modified, circulated within closed groups, and used as a basis for further manipulation. Screenshots of threats may be dismissed as insufficiently reliable or probative, particularly where perpetrators deny authorship or claim that content has already been removed. Courts and investigators are asked, in effect, to adjudicate harm in a domain where the boundary between evidence and fabrication is itself contested. Without robust, widely accessible tools for authentication and provenance, there is a risk that institutional scepticism about synthetic media will be displaced onto victims, reinforcing the epistemic injustices discussed above.

These enforcement challenges do not, on their own, justify a retreat from criminalisation or a shift of responsibility onto victims. They do, however, underscore the limitations of an approach that relies exclusively on ex post investigation and individualised prosecution. Legal responses must be recalibrated towards shaping the technological and economic conditions under which synthetic coercion can operate, while ensuring that investigative agencies are equipped—and required—to treat deepfake sextortion as a serious form of sexual and economic exploitation.

5.4. Towards a UK Accountability Model for Synthetic Coercion

The foregoing analysis suggests that incremental adjustments to existing offences and platform duties will not be sufficient. A coherent UK response requires an integrated accountability model built around three pillars: substantive recognition of synthetic sexual harm; recalibrated platform responsibilities under the OSA; and accessible, victim-centred remedies.

(a) 

Substantive recognition of synthetic sexual harm

The first pillar concerns the definition of the wrong itself. The Law Commission’s reclassification of intimate image abuse as a sexual autonomy offence provides an important starting point (Law Commission 2022). This logic should be extended explicitly to synthetic imagery. A base offence should criminalise the intentional creation, alteration, possession for the purpose of sharing, sharing, or threatening to share sexually explicit synthetic images of an identifiable person where the defendant knows that the person does not consent, is reckless as to whether they consent, or lacks a reasonable belief in consent. The offence should be motive-independent and should apply irrespective of whether dissemination ultimately occurs, so that threat-stage coercion is captured as a completed wrong rather than treated as an inchoate precursor. Aggravated forms should capture conduct involving extortion, coercive control, repeat targeting, or children.

An offence carrying significant stigma and potential custody cannot dispense with culpability without raising acute questions of proportionality and fair trial under Article 6 ECHR and without placing strain on the legality principle in Article 7 ECHR. The proposed structure therefore does not proceed on a strict-liability basis. It confines criminalisation to intentional conduct combined with fault as to consent (knowledge or recklessness, with an explicit reasonable-belief route), preserving the requirement that culpability be proved beyond reasonable doubt and maintaining a foreseeable boundary between lawful expression and unlawful synthetic sexual appropriation.

Northern Ireland Human Rights Commission proposals for deepfake reforms, which emphasise the need for motive-independent base offences and highlight the human rights implications of fragmented protection, are instructive in this regard (Northern Ireland Human Rights Commission 2025). A similar approach across England, Wales and Northern Ireland would reduce uneven protections contingent on geography. It would also reflect the reality that, for victims, the core wrong lies in the non-consensual appropriation and sexualisation of their image and the use of that representation as a tool of control—not only in the act of publication.

Substantive reform should be complemented by the creation of a statutory civil cause of action for image-based abuse, encompassing synthetic sexual harm. Such a tort would enable victims to seek injunctions and damages without depending on state prosecution or squeezing their claims into ill-fitting privacy, harassment or defamation doctrines. The threshold should be framed around non-consensual use or threatened use of sexualised representations, coupled with proof of foreseeable distress or interference with autonomy, rather than around authenticity or demonstrable reputational loss.

(b) 

Recalibrated platform duties under the Online Safety Act

The second pillar requires a more ambitious use of the OSA’s regulatory machinery. Existing duties of care should be specified, through Ofcom’s codes of practice, in ways that respond directly to the dynamics of deepfake sextortion. At minimum, regulated services—including social media platforms, search engines, pornography sites and app stores—should be required to:

• Conduct granular risk assessments focused specifically on intimate image abuse and synthetic sexual harm, including threat-based abuse rather than only published content;
• Implement safety-by-design measures that restrict or prohibit on-service tools and functionalities, and limit distribution or promotion on the service, where the primary or predominant use is the generation of non-consensual sexually explicit imagery;
• Adopt clear, rapid-response procedures for sextortion reports, including preservation of evidence, immediate restriction of relevant accounts, and cooperation with law enforcement; and
• Maintain auditable records of their handling of intimate image abuse cases, subject to Ofcom oversight and sanctions for systemic non-compliance.

The Children’s Commissioner’s recommendations on nudification tools point to the types of design and distribution controls required (Children’s Commissioner for England 2025). App stores and search providers should be explicitly brought within the ambit of intimate-image risk duties, reflecting their role as gateways to deepfake generators. Ofcom’s guidance on deceptive deepfakes could be developed into binding expectations for watermarking, provenance information, and the interoperability of detection tools, while recognising the need for safeguards against over-removal and discriminatory error. Crucially, platform obligations should not be limited to the handling of content once it is public. Codes of practice should require services to identify and mitigate patterns indicative of sextortion—such as multiple users reporting similar threats from common accounts, or use of particular templates and payment demands—even where the underlying material has not been posted. That approach would align system design with the temporal reality of harm, which arises at the moment of threat.

(c) 

Victim-centred redress and institutional coordination

The third pillar concerns remedies and institutional design. Victims of deepfake sextortion currently navigate a complex mixture of criminal reporting, civil claims, and informal complaints to platforms, often without legal advice or specialist support. A coherent accountability model would create clearer, faster pathways and ensure that institutions are coordinated rather than siloed.

One option would be a fast-track protective order regime, modelled loosely on non-molestation orders, enabling courts to prohibit specified conduct relating to synthetic sexual imagery (including creation, possession, sharing and threatening to share) on an urgent basis. Such orders could be available on application by victims or on the motion of the police, with breach criminalised by statute as part of the protective order scheme. Their function would be preventive, addressing temporal misalignment by recognising the threat itself as a legally cognisable form of harm. In parallel, there is a strong case for designating specialist “trusted reporter” organisations—such as NGOs working on image-based abuse—to interface with platforms and Ofcom on behalf of victims. These bodies could coordinate takedown requests, liaise with law enforcement, and assist in preserving evidence, reducing the burden on individuals who may be dealing with acute distress.

Institutionally, responsibility for synthetic sexual abuse could be embedded within existing strategies on violence against women and girls and online abuse, rather than treated as a niche cybercrime issue. This would entail explicit recognition, within Home Office and Ministry of Justice policies, that deepfake sextortion is a form of gender-based violence and economic exploitation, and corresponding commitments to training, resourcing and monitoring. Coordination between Ofcom, policing bodies and victim-support services would need to be formalised, with shared protocols and data-sharing arrangements that respect privacy while enabling pattern recognition and systemic interventions.

5.5. Embedding Recognition: Towards a Jurisprudence of Plausible Fictions

The preceding analysis exposes a persistent doctrinal misfit: authenticity continues to operate as an implicit threshold for legal recognition in contexts where the wrong consists of coercive deployment of a plausible fabrication. The law has long recognised that threats, attempts and conspiracies can be wrongful even when the envisaged harm does not materialise; liability attaches to the targeting of protected interests and the imposition of coercive constraint, not only to completed injury. Deepfake sextortion requires an equivalent adjustment in the treatment of image-based abuse. The fact that a sexualised representation is fabricated rather than authentic does not diminish its capacity to operate as leverage against an identifiable person, or to generate threat-stage harms that crystallise prior to any dissemination.

A jurisprudence of plausible fictions would shift legal attention from the factual status of the depiction to the mechanism and effects of its deployment. “Plausible fictions” denotes fabricated representations capable of being treated as credible by relevant audiences and used, or threatened to be used, as leverage against an identifiable target. Authenticity would remain relevant to evidential assessment and to the classification of wrongdoing, but it would not function as a gatekeeping criterion for recognition. The central questions would be: did the defendant appropriate or fabricate a sexualised representation of an identifiable person without consent or without a reasonable belief in consent; and was that representation used, or threatened to be used, in a manner that foreseeably undermined the person’s autonomy, dignity or security. The focus is therefore on non-consensual appropriation coupled with coercive deployment, including threats and conditional demands.

Any such reorientation must be compatible with Article 10 ECHR, including the Strasbourg Court’s insistence on the special protection afforded to political expression and satire, even where it is provocative or sexually explicit. The framework should not treat sexualised depictions of public figures as presumptively unlawful. The limiting principle is functional and conduct-based: deepfake sextortion is constituted by coercive deployment—threat, conditional demand, intimidation, or leverage—using a fabricated sexual depiction to constrain another’s options. That concerns interference with autonomy, not the regulation of political viewpoint. The regulatory implications under the OSA should likewise be directed to coercive patterns and unlawful threats, while maintaining safeguards against over-removal, contestable classification, and discriminatory error.

Embedding this perspective within the law of England and Wales and Northern Ireland would bring several benefits. It would align substantive offences and civil causes of action with the harm analysis developed in earlier sections, reduce institutional discounting premised on the absence of “real” images, and clarify that the wrong is not defined by communicative content alone but by non-consensual appropriation of an identifiable person’s likeness and its deployment as coercive leverage. It would also provide a clearer normative basis for platform regulation under the OSA, supporting specification of safety-by-design duties and oversight where services predictably facilitate the coercive use of synthetic media. Those regulatory measures should be formulated and applied in a manner that preserves lawful expression, including political satire, through contextual assessment and accessible routes to challenge removal or restriction decisions. Most importantly, such a jurisprudence would reallocate the practical burdens of detection, reporting, and evidential preservation away from victims and towards those who create, distribute, and weaponise synthetic imagery, and towards the services whose systems enable its circulation. Recognising plausible fictions as a proper object of legal concern is a necessary step in ensuring that institutional and regulatory design choices do not undermine the autonomy and dignity of those most exposed to synthetic coercion.

6. Conclusions: Seeing the Unseen Harm

Deepfake sextortion exposes a persistent doctrinal uncertainty in the law of England and Wales and Northern Ireland: whether threat-stage coercion implemented through fabricated sexual depictions is treated as a legally salient wrong, or whether recognition remains contingently tied to authenticity, dissemination, and repeated conduct. The analysis demonstrates that existing criminal, civil, and regulatory routes respond only unevenly because fabrication is still handled as an evidential complication rather than as the mechanism through which coercive constraint is imposed. Doctrinal coherence depends on treating non-consensual representational appropriation and coercive deployment, rather than authenticity or eventual dissemination, as the organising features of analysis.

The analysis isolates deepfake sextortion as coercion through falsity: a completed wrong constituted by the non-consensual fabrication or appropriation of an identifiable person’s sexualised depiction and its use, or threatened use, as leverage to constrain options, induce compliance, extract payment, or enforce silence. This frame integrates the paper’s conceptual claims about threat-stage, anticipatory and epistemic injury with its doctrinal evaluation of offence architecture, civil predicates, and the service-based structure of the Online Safety Act 2023. The doctrinal difficulty is not a general inability to address threats; it is the mismatch between the modalities of coercion in this setting—fabricated depictions functioning as putative evidence, rapid circulation pathways, and credibility assessment under evidential uncertainty—and legal triggers that presuppose different predicates for harm, responsibility, and intervention.

Three recurrent mischaracterisations have been corrected, each affecting classification and outcomes. Fabrication does not reduce legal salience where plausibility supplies coercive leverage; the operative question concerns constraint imposed through threatened deployment. Deepfake sextortion cannot be treated primarily as a post-publication content problem; in many instances the harm crystallises at the point of threat, including through private transmission and imminent upload. Residual civil and communications-law routes cannot be assumed to supply functional equivalents where they depend on predicates such as publication, private facts, or repeated conduct, and therefore do not reliably deliver timely prevention or stable recognition of the wrong.

These findings entail concrete implications for decision-makers. Courts and prosecutors should treat the fabricated depiction as the vehicle of menace and direct analysis to threat, conditional demand, and coercive constraint, rather than defaulting to categories that recast the conduct as mere offensiveness or evidential novelty. OSA implementation should be assessed in terms of what the statute requires and permits: service-based risk management anchored in duties concerning illegal harms, capable of supporting rigorous specification of expectations responsive to sextortion dynamics (rapid response, evidence preservation, and mitigation of predictable pathways), while remaining bounded by legality, proportionality, and safeguards against over-removal and discriminatory error. Article 10 ECHR constrains both criminalisation and regulatory practice: the limiting principle must remain conduct-based coercive deployment, not the mere existence of sexualised depiction, including in political contexts.

The policy and doctrinal recommendations follow directly. First, intimate-image offences should be recalibrated to capture synthetic sexual depictions through a motive-independent base offence framed around non-consensual appropriation and coercive deployment, with a defined fault element and an explicit reasonable-belief route, and aggravated forms for extortionary demands and coercive control. Secondly, OSA codes and guidance should specify measures that address threat-stage sextortion dynamics and recurring weaponisation within service design, coupled with transparent safeguards and reviewable processes for contested removals. Thirdly, victim-centred procedural routes should be strengthened through fast-track protective mechanisms and formalised coordination between reporting, platform processes, and enforcement, so that protection is not contingent on post-publication escalation.

Limits remain. Difficult classification problems at the margin between coercive deployment and protected expression will remain, particularly where sexualised depictions of public figures are advanced as commentary rather than deployed as leverage through threats, conditional demands, or intimidation. Attribution problems in encrypted, pseudonymous, or cross-border settings will continue to constrain enforcement and prevention. Service-based regulation cannot reach every channel through which threats are communicated, and institutional capacity and technical error rates in detection and classification will continue to shape outcomes even under improved doctrinal framing.

Recent developments elsewhere, such as the United States’ TAKE IT DOWN Act 2025—the first explicit federal statute prohibiting the non-consensual publication of intimate images, including AI-generated deepfakes, and requiring online platforms to remove notified content within 48 h—underline the emerging expectation that synthetic sexual imagery be treated on a par with authentic material in both criminalisation and takedown obligations (Congressional Research Service 2025). The jurisprudence of plausible fictions developed in this paper enables more predictable classification of conduct, more principled allocation of responsibility between offenders, services, and institutions, and bounded discretion compatible with legality and freedom of expression. That matters because uncertainty in this domain is operational: it determines whether threat-stage coercion is recognised in time to prevent escalation and whether institutional decisions are taken on doctrine-consistent, reviewable grounds.