Desire in Code: Legal Perspectives on Sex Robots and Consent

Abstract

This paper explores the intersection of sex robots, rape, and sexual consent, examining their legal and normative challenges. Designed for sexual interaction—primarily for a cisgender male, heterosexual market—sex robots risk reinforcing women’s subordination and oppression. Their ability to simulate sexual consent further complicates their societal and legal positioning. While some advocate prohibition, this paper examines regulatory alternatives that embed affirmative consent in design. In particular, it considers the potential of the AI Act and its provision on codes of conduct to guide such development.

I. Introduction

“Lucrece,” quoth he, ‘this night I must enjoy thee:

If thou deny, then force must work my way, For in thy bed I purpose to destroy thee:

That done, some worthless slave of thine I’ll slay, To kill thine honour with thy life’s decay. ¹

William Shakespeare, The Rape of Lucrece (1594)

Although the narrative of Lucrece’s rape dates back to the sixth century BC and was later immortalized by William Shakespeare in the late 16th century AD, its resemblance to contemporary news remains undeniable. Despite challenges in collecting and comparing global rape statistics due to variations in legislation and methodologies, ² more women are speaking out and uniting to combat this pervasive issue. Since the 1970s, women’s rights groups have advocated for recognizing rape as gender-based violence, challenging victim-blaming attitudes, and promoting consent-based definitions in legal frameworks. ³ While progress has been made, particularly in the European Union, ⁴ substantial obstacles persist. Notably, the European Parliament and the Council have failed to reach a political agreement on criminalizing rape based on consent, ⁵ excluding this harmonization from Directive (EU) 2024/1385 on violence against women and domestic violence.

Simultaneously, the socio-legal discourse on rape has increasingly expanded beyond human interactions to address its representation and mediation. In the 2010s, a significant debate emerged over the morality and harm of video games centered around rape. This discussion gained prominence when European activists, politicians, and scholars condemned the Japanese video game RapeLay, depicting the sexual harassment and rape of a woman and her underage daughters. Their efforts led to restrictions on its sale and production. ⁶ Similarly, in 2015, the British Parliament broadened the criminalization of extreme pornography to include depictions of rape, citing its potential to normalize sexual violence and reinforce women’s subordination. ⁷ More recently, socio-legal debates have turned to emerging technologies in mediating rape, such as remotely controlled sex toys, which risk enabling rape by deception through misrepresentation of a partner’s identity or the sexual act. ⁸ Incidents of “rape” within the metaverse have also increased, prompting further investigations. ⁹

In this context, the debate on representing and mediating rape has expanded to include sex robots and the potential for their prohibition. These human-like machines, designed to deceptively simulate emotional and sexual interaction, raise concerns about perpetuating discrimination and violence against women, ¹⁰ echoing earlier debates about rape representation in video games and pornography. However, due to their physical embodiment, discussions around sex robots also intersect with concerns about whether remotely controlled sex toys and the metaverse can mediate rape, as all blur virtual and physical boundaries and may amplify or create new sexualized harms against women. While some authors imagine sex robots that can reject sexual advances, ¹¹ others document their real-world existence, ¹² reinforcing the call for the outright ban made by the Campaign against Sex Robots in 2015. ¹³ Although this issue has attracted attention in media and academic circles, ¹⁴ the legal qualification of “rape” in user-sex robot interactions remains underexplored. Particularly, little attention has been paid to how the cultural harm these human-like, deceptive machines are said to reproduce—by embodying stereotypical representations of female appearance and behavior and fostering a socio-cultural milieu that normalizes and legitimizes discrimination and violence against women—intersect with their capacity to simulate consent, including its provision, withdrawal, or refusal. This gap is critical amid the ongoing race to regulate emerging technologies, such as artificial intelligence (AI), in the European Union, without fully incorporating a gender-sensitive and responsive approach. ¹⁵

Accordingly, this paper examines how sex robots challenge existing legal and normative understandings of “rape” and “sexual consent,” by unpacking their relationship to socio-legal interpretations, interrogating the implications of simulated consent, and engaging with feminist literature on rape culture and online and technology-facilitated violence. This analysis is situated within broader debates on women’s subordination and AI governance, where the regulatory challenge extends beyond identifying harm to recognizing how such technologies encode and perpetuate structural power relations. It foregrounds how sex robots may reinforce normative hierarchies that continue to compromise the dignity, autonomy, and equality of women as historically oppressed within socio-legal frameworks. While acknowledging and building on feminist and legal critiques of related societal practices and institutions, like pornography, video games, sex toys, and the metaverse, this paper treats sex robots as a distinct site of cultural harm, owing to their unique fusion of physical embodiment with deceptive, gendered, and sexualized forms of interaction.

The paper is structured as follows. Section 2 sheds light on sex robots, explaining their nature and highlighting the societal concerns they raise. Section 3 situates sex robots within different interpretations of sexual consent, exploring how their programming—whether to refuse or implicitly offer consent—intersects with theories of rape culture and online and technology-facilitated violence. Subsequently, Section 4 focuses on regulation, advocating for alternative designs of sex robots, particularly those incorporating affirmative consent programming, over prohibitionist approaches to address the issues of rape culture and online and technology-facilitated violence that sex robots can perpetuate. The section also engages with Regulation (EU) 2024/1689 (commonly referred to as the AI Act or the AIA) using Article 95 on codes of conduct as a case study to explore voluntary regulation of sex robots grounded in principles of affirmative consent within the European Union. The paper concludes with a summary of the findings.

II. Conceptual Foundations of Sex Robots

2. The Making of Sex Robots: Historical Precedents and Modern Design

The very idea of creating artificial partners is not new, often referenced through the ancient myth of Pygmalion to sexual aids like the so-called dames de voyage and “Dutch wives” carried by sailors from 16th century and rudimentary sex dolls provided to soldiers in World War II. ¹⁶ In the 20th century, advancements in materials paved the way for significant evolution in sex doll manufacturing. Precisely in the 1970s, the introduction of vinyl, latex, and silicone facilitated the creation of the first inflatable sex doll, offering a life-size representation of a female figure with penetrable orifices. By the 1990s, realistic sex dolls began to emulate the feel and texture of human skin, incorporating features like an articulated steel skeleton for more authentic postures. The progression continued into the 2000s, witnessing the integration of robotics and AI. ¹⁷ Currently, generative AI (GenAI) is widely regarded as set to transform the industry, enhancing sex robots’ ability to adapt to users’ preferences and behaviors. ¹⁸

Despite the existence of multiple manufacturers and production lines, the sex robot industry remains marginally mapped, ¹⁹ with Realbotix dominating public narratives and media attention. ²⁰ Nonetheless, sex robots are broadly characterized by their human-like features, like realistic hair, articulated limbs, and anatomically detailed genitals. Increasingly, these machines are equipped with exoskeletons that enable movement, walking, and untethered mobility, powered by internal batteries. Sex robots are also embedded with external and internal sensors that allow them to respond to user actions, simulate personality traits, engage in conversations, and produce sounds indicative of sexual arousal. Some models even exhibit a degree of autonomy, capable of mimicking human behavior, displaying spontaneous sociability, and adapting to interactions through learning, often augmented by emotion recognition systems. ²¹

A defining characteristic of sex robots is their hyper-realistic, human-like design, which extends beyond physical resemblance to encompass behavioral traits. While these machines are often marketed as highly customizable, ²² this opportunity has often resulted in the creation of women-like machines characterized by stereotypical physical features, including accentuated breasts, prominent buttocks, and slim waists. Similarly, the available personalities frequently reflect conventional and reductive fantasies of the “ideal” female, including traits such as being annoying, jealous, moody, or frigid. ²³ Racialized attributes and personal characteristics like age further illustrate how sex robot design is deeply intertwined with societal stereotypes and structural inequalities. ²⁴ Furthermore, some sex robots have been modeled after real individuals, particularly women, without their consent. ²⁵

Beyond their esthetic and behavioral characteristics, sex robots are often designed to create the illusion of possessing human-like capabilities that they do not actually have. Developers deliberately obscure their artificial nature, blurring the boundary between human and machine to enhance user engagement. ²⁶ While this design strategy is not unique to sex robots and is commonly employed in social robotics to enhance human-machine interaction (HMI) and facilitate societal integration, ²⁷ the notion of “full deception,” where robots might be mistaken for humans on both conscious and subconscious levels, remains highly contentious. ²⁸

2.1. Sex Robots, Societal Implications, and Gender Dynamics

In the literature, sex robots are often regarded as tools for self-empowerment and expression within the sexual and intimate spheres. They are considered potential partners and could even play a role in sex work, diversifying intimate experiences, circumventing criminal prosecution, and addressing issues such as human trafficking and the spread of sexually transmitted diseases. ²⁹ Furthermore, sex robots could support marginalized individuals, such as the elderly or people with disabilities, in fulfilling their sexual desires, ³⁰ while also offering valuable tools for sexual and emotional education. ³¹ Advocates have even suggested that their customizable design could promote greater acceptance of diverse gender identities and sexual orientations. ³² However, numerous concerns have been raised. While some scholars emphasize potential risks related to user degradation and data protection, ³³ most critical perspectives focus specifically on the gendered nature and implications of this human-like, deceptive technology, which embodies stereotypical representations of female appearance and behavior, thereby contributing to the normalization and legitimization of gender-based subordination and sexual violence.

In 2015, Kathleen Richardson from Leicester De Montfort University initiated the Campaign against Sex Robots, ³⁴ later renamed the Campaign against Porn Robots. ³⁵ Her campaign draws comparisons between sex robotics, prostitution, and pornography. Richardson argues that technology is inherently shaped by values and reflects the prevailing power imbalances between women and men. She claims that sex robots, particularly those designed to mimic women, perpetuate these gender disparities by reinforcing stereotypical portrayals of female appearance and behavior found in mainstream pornography. Furthermore, their design fosters an instrumental relationship, where the machine is intended solely to fulfill the sexual desires of the user without offering any emotional reciprocation, mirroring the dynamics of prostitution. Richardson concludes that by offering a distorted and misleading depiction of female behavior and appearance, sex robots contribute to women’s sexual commodification, reinforcing the notion that both sex robots and women are objects to be used for male sexual pleasure. ³⁶

Several authors have raised similar concerns, arguing that the stereotypical and deceptive design of women-like sex robots perpetuates social discrimination and sexual violence that women face, unlike men, due to their gender. ³⁷ At the core of these concerns is the idea that sex robots reduce women to interchangeable objects, treated as substitutes or enhanced versions of real women, owing to their human-like appearance and constant availability. Research further suggests that some users non-consensually create sex robots based on their partners or former partners, perceiving them as superior, partly due to their lack of independent thought. ³⁸ More broadly, sex robots are seen as contributing to a “rape culture” by normalizing and legitimizing sexual violence against women, while also undermining the concept of sexual consent, ³⁹ a theme that is explored in more detail below.

While these critiques highlight urgent socio-legal concerns, the extent to which sex robots influence societal attitudes and behaviors—whether reinforcing harmful norms or potentially offering alternative forms of self-empowerment and expression—remains underexplored. Empirical research on their impact, both at an individual level and societal level, is still limited, ⁴⁰ leaving open questions about how these technologies shape, and are shaped by, broader social and legal frameworks. Yet, this tension is far from new, and the pursuit of empirical evidence should not be assumed to offer a definitive resolution. Debates surrounding sex robots echo longstanding feminist disagreements over the social construction and regulatory disciplining of gender and sexuality, disagreements already evident in the context of pornography. ⁴¹ There, empirical studies have generated divergent conclusions, variously claiming that pornography commodifies sex or fosters sexual autonomy. ⁴² These contradictions highlight that evidence alone cannot settle normative conflicts, and that legal responses are ultimately shaped by value-laden judgments. Normative frameworks delineate what is considered socially permissible, often through rigid classifications and sexual double standards, functioning as disciplinary mechanisms that, when inscribed in law, deepen social exclusion and diminish dignity. ⁴³ It is therefore essential to interrogate not only the technology itself, but also the normative structures within which it is embedded. Binary framings of sex robots—or of societal institutions more broadly—as either empowering or harmful risk oversimplifying the complex, intersecting dynamics through which gender, sexuality, power, and regulation coalesce, and through which both law and technology may perpetuate systems of privilege and oppression.

III. Sex Robots and the Meaning of Rape

3. Imagining and Experiencing Rape in User–Sex Robot Interactions

In Robots, Rape, and Representation, Robert Sparrow challenges readers to imagine a sex robot resembling a woman, programmed to display sexual desire and pleasure while explicitly “consenting” to sex. ⁴⁴ However, Sparrow extends the thought experiment by introducing a scenario in which the machine could reject sexual advances, verbally refusing and physically retreating. In this case, the robot might cry out phrases like “Stop it, you are raping me. Stop raping me. He’s raping me” ⁴⁵ while trying to halt the act. Similarly, Sinziana Gutiu critiques the nature of consent in sex robots, which are designed to comply with any sexual interaction initiated by their users. For Gutiu, this predetermined compliance strips the concept of consent of its significance in HMIs. ⁴⁶ The works of Sparrow and Gutiu may feel less like speculative fiction than they first appear, especially when considered alongside Ellen M. Kaufman’s qualitative study on sexual consent in user–sex robot interaction. ⁴⁷

Kaufman examines the prototype “Frigid Farrah,” a sex robot programmed to resist “inappropriate” or “unwanted” physical contact by rejecting and opposing user advances. While some authors find the notion of a woman-like machine programmed to refuse sexual engagement deeply troubling, and others dismiss it as an exaggerated or possibly non-existent prototype, ⁴⁸ the manufacturer justifies this feature as a tool for teaching intimacy, framing it—at most—as a reflection of a “reserved and shy personality.” ⁴⁹ Expanding on this, Kaufman critically explores how sexual consent is framed and operationalized in sex robots. Her findings reveal that many users view consent as an undesirable trait, perceiving it as an obstacle to the unrestrained and one-sided gratification they seek. ⁵⁰ Complementary research by Anastasiia D. Grigoreva et al. reveals a growing trend of blaming and penalizing those who commit sexual violence against human-like robots. ⁵¹ This research highlights not only the potential role of sexual consent in HMIs but also the evolving perception of moral accountability when such boundaries are violated.

In this context, the lack of a clear definition of sexual consent in user–sex robot interactions remains a significant gap. Furthermore, it is unclear how the absence of consent—or its failure to be explicitly expressed—might align with the legal classification of rape.

3.1. Sexual Consent Between Humans: Law, Norms, and Power

In legal literature, sexual consent is often described as the “magic” that changes the moral and legal dynamics between individuals. ⁵² It makes certain sexual acts permissible where they would otherwise be prohibited, while also redefining rights and obligations between the parties involved and the State. By consenting to sexual advances, an individual authorizes previously restricted actions and simultaneously limits the State’s role in safeguarding their autonomy. ⁵³ Nonetheless, although scholars generally agree that consent distinguishes permissible from impermissible sex and prevents sexual commodification, ⁵⁴ there is little agreement on how to define its boundaries. For further analysis, Peter Westen’ distinction between the factual and legal dimensions of sexual consent offers a robust framework, enriched by contributions from various disciplines. ⁵⁵

Westen highlights that, factually, sexual consent encompasses both subjective and objective dimensions. This means an individual may internally feel willing to engage in a sexual act and/or outwardly express this willingness through words or actions. ⁵⁶ The factual dimension reflects how people navigate mutual agreement in everyday interactions, acknowledging that consent is fluid and must be specific to each act within a sexual encounter. ⁵⁷ However, societal beliefs about gender and sexuality often shape how individuals negotiate consent, with such norms unconsciously influencing their choices and actions. ⁵⁸ Consequently, in legal contexts, factual interpretations of consent have led to controversial rulings. For instance, in the so-called “Wolf Pack” case (2018), a Spanish court interpreted the victim’s passive behavior—closing her eyes and enduring the assault—as implicit consent, leading to an initial acquittal for gang rape. ⁵⁹

Building on the factual dimensions of sexual consent, Westen argues that consent holds legal validity when its attitudinal and expressive elements align with a community’s normative standards, typically defined by competence, freedom, and knowledge. ⁶⁰ While the precise thresholds for these criteria remain contested, ⁶¹ they provide a useful analytical framework. Briefly, competence first refers to the cognitive and emotional capacities needed to grasp the nature of consent and assess its risks and benefits. ⁶² These must be complemented by social competence, often shaped by life experience. ⁶³ Second, knowledge requires individuals to be adequately informed to make rational decisions, including understanding the circumstances and consequences of their actions. ⁶⁴ Finally, freedom means the ability to give consent without coercion, manipulation, or the fear of significant repercussions, with power imbalances—such as those related to gender or age—potentially undermining it. ⁶⁵

The gap between theory and practice is considerable, and these theorizations of consent often fall short in real-world application. While awareness of consent requirements may seem widespread, recognizing a partner’s willingness to engage in sexual activity remains complex. Consent is communicated and interpreted in diverse ways, shaped by personal traits and situational contexts. Research shows that women, regardless of the interaction being heterosexual or same sex, are more likely to use verbal signals, whereas men often rely on nonverbal cues. ⁶⁶ Explicit communication is also more common in established relationships and for penetrative sex than for other forms of intimacy, like kissing. ⁶⁷ This variability makes defining sexual consent in a single legal category particularly challenging, often leading to over- or under-inclusive applications. ⁶⁸ Beyond definitional challenges, the formalization of consent in law has drawn criticism for potentially expanding State intervention in private life while reinforcing narratives that portray women as emotionally vulnerable in ways that may be repressive and sex-negative rather than fostering meaningful social change. ⁶⁹ Evidentiary hurdles further complicate matters, as conflicting narratives can obscure fact-finding, and normative assessments remain deeply contested. ⁷⁰ Moreover, gender and sexual stereotypes continue to shape courtroom dynamics, leading judges to rely on biases rather than rigorous legal reasoning.

Overall, the social construction and legal definition of sexual consent remain fraught with complexity. While consent plays a crucial role, addressing these challenges requires broader societal change, starting with greater access to accurate sexual information to destigmatize discussions about sex and promote pleasurable, rather than merely tolerable, experiences. ⁷¹ Some progress is evident in contexts like sex work and bondage, discipline, sadism, and masochism (BDSM), where power asymmetries are more openly acknowledged, and consent is central without diminishing intimacy or eroticism. ⁷² However, for consent to be truly meaningful, efforts must also focus on dismantling systemic inequalities affecting women and other marginalized groups. ⁷³ Its operationalization should thus intersect with feminist debate, often framed around “no means no” and “yes means yes” models. The former assumes individuals can actively refuse unwanted advances, placing responsibility on men to respect clear refusals. ⁷⁴ Conversely, affirmative consent shifts the burden onto men to ensure mutual desire and explicit agreement, acknowledging power imbalances that may inhibit women from expressing non-consent and reducing the evidentiary burden on victims in legal disputes. ⁷⁵

These ongoing debates around the meaning and operationalization of consent, particularly in contexts shaped by structural inequalities and respect for sexual autonomy, primarily concern human interactions. Yet, as illustrated by legal and ethical narratives on sex with animals, often characterized as “a-consensual interactions,” ⁷⁶ questions arise as to how existing understandings of consent might apply to sex robots. Designed to simulate sexual encounters while often disregarding the principles of consent, sex robots necessitate a closer examination of their normative implications for gendered and sexualized relations within society.

3.2. The Symbolic Role of Consent in User–Sex Robot Interaction

As discussed in Section 3.1, sex robot prototypes like “Frigid Farrah” are designed to simulate sexual refusal, allowing users to rehearse coercive scenarios as if interacting with a woman, without real-world consequences. More broadly, the implicit consent model in most sex robots diverges from legitimate understandings of factual—especially affirmative—consent, potentially reinforcing non-consensual sexual behavior. ⁷⁷ These designs likely reflect gendered power dynamics, including stereotyping, biased treatment, and emotional responses, in HMIs. ⁷⁸ While fictional debates continue whether sex robots can understand consent or experience rape, ⁷⁹ the issue is not their suffering but their symbolic impact. ⁸⁰ This is comparable to current discussions about the simulated killing of anthropomorphic robots designed for social interaction, where calls for their protection are motivated less by concern for the machines themselves and more by the desire to discourage behaviors that would be harmful if directed toward human beings. ⁸¹ Similarly, although sex robots neither consent nor experience harm as humans do, their programmed and deceptive responses to sexual advances can distort the understanding of sexual consent within gendered relations, while misleading users about the artificial nature of their interactions. Accordingly, consent could become a necessary criterion for legitimizing encounters with human-like machines, ⁸² aligning with relevant legislation and literature.

Although sex robots lack the competence, knowledge, and freedom for legal consent, they can still simulate refusal of sexual advances and provide clear signals of consent. Their design allows them to replicate factual consent, particularly in its expressive dimension. However, non-consensual interactions with human-like machines differ fundamentally from those between people as their interactions rely on representations rather than direct engagement. While sexual consent in human relationships attempts to safeguard sexual autonomy and typically requires criteria such as competence, knowledge, and freedom, consent in human–sex robot interactions serves a more symbolic role. Since users often perceive sex robots as human-like, the sex robot’s responses and the user’s reactions shape understandings of sexual consent in gendered relationships. ⁸³

Accordingly, sex robots should be designed to explicitly communicate consent, such as through verbal affirmation or other behaviors that simulate positive desire and willingness. This would prevent users from misinterpreting their accommodating design as actual consent and mitigate the risk of users deriving satisfaction from programmed resistance to unilateral advances. While this cannot substitute for the complexity of human interactions—where consent is often fluid, situational, and influenced by social scripts—it can challenge harmful assumptions, promote the desirability of affirmative cues, and support broader efforts to normalize consent as a necessary, dynamic, and mutual process. This design orientation also resonates with the said shift toward sex positivity and contributes to the development of factual and legal understandings of consent that are sensitive and responsive to gendered and sexual inequalities within society.

3.3. The Non-Consensual Design of Sex Robot as Rape Culture and Online and Technology-Facilitated Violence Against Women

Given the symbolic nature of consent in user-sex robot interactions, discussions surrounding the rape of a sex robot should be framed within the context of feminist theories on “rape culture,” ⁸⁴ which emerged in the 1970s to address the widespread normalization of gender-based and intersectional violence. ⁸⁵ Rape culture highlights how hegemonic masculinity is reinforced through language, media, and, more recently, through the design and deployment of new technologies like sexual violence and harassment in the metaverse. ⁸⁶ In this light, the term “rape” can serve to convey that non-consensual design and interactions with sex robots perpetuate women’s sexual commodification. This interpretation acknowledges the broader societal implications of such gendered and deceptive technologies, fulfilling its expressive function without conflating the more immediate impact of sexual violence.

In addition, this interpretation suggests that non-consensual design and interaction with a sex robot could represent a novel form of online and technology-facilitated violence against women. While various definitions of such violence exist, ⁸⁷ the Platform of Independent Expert Mechanisms on Discrimination and Violence against Women outlines four broad categories based on the role of technology. The first category addresses harassment and violence facilitated by specific technologies, such as spyware or tracking devices used in intimate partner violence. The second involves abuse that occurs and is exacerbated online, such as the non-consensual sharing of intimate images. The third encompasses emerging forms of abuse enabled by technology, like sexually explicit digital forgeries. Finally, the fourth category covers the use of information and communication technologies (ICTs) to enable violence, with social media playing a significant role in disseminating misogynistic content. ⁸⁸

Although currently limited in scale and supported by scarce evidence, ⁸⁹ sex robots may give rise to forms of harm that resemble the third and fourth categories of online and technology-facilitated violence. One such form of sexualized and gendered harm arises when sex robots are modeled on identifiable women without their consent. These cases violate their sexual autonomy and dignity, ⁹⁰ producing a specific kind of harm that was not possible prior to the production of this technology. Beyond these more concrete instances, most sex robot prototypes can also be understood as technologies that facilitate violence in a more structural way—not by targeting individuals, but by contributing to harm directed at women as a gendered group. Critics argue their gendered and deceptive design reinforces women’s sexual commodification, distorts consent and intimacy, and perpetuates rape culture. Their hyper-realistic appearance and potential for non-consensual programming blur the line between technological and physical abuse, enabling and normalizing harmful dynamics in human relationships. ⁹¹ Sex robots are believed to cause broader cultural harm ⁹² by desensitizing users to sexual autonomy and gender equality, potentially fostering attitudes that translate into real-world subordination and oppression of women. Precisely, they may encourage the internalization of harmful norms that trivialize the relevance of consent or treat women as sexually available by default. These attitudes are expected to extend beyond the realm of HMI, influencing users’ expectations and behaviors in their interpersonal relationships. In this way, sex robots are believed to reinforce a socio-cultural milieu that normalizes and institutionalizes women’s subordination and oppression. ⁹³ The impact of such cultural harm is not seen in isolation; it is argued that it operates synergistically with other cultural practices and institutions—including pornography, prostitution, and even sexist or violent video games—collectively shaping societal attitudes that perpetuate gender-based discrimination and violence. ⁹⁴

More broadly, sex robot engagement shares characteristics with online and technology-facilitated violence against women, often considered gender-based. ⁹⁵ As noted in Section 2.1, this gendered nature is evident in their stereotypically female design and predominant use by cisgender heterosexual men. Furthermore, online and technology-facilitated violence aligns with a broader continuum of abuse that women have historically faced, with interconnected and recurring patterns of harm. ⁹⁶ Sex robots contribute to this continuum by normalizing sexist attitudes, reinforcing stereotypes, and perpetuating sexual commodification. In addition, online and technology-facilitated violence—including the design and use of sex robots—can affect women in complex and varied ways. ⁹⁷ These human-like machines often carry gendered implications that intersect with other forms of social oppression, such as race and age, potentially magnifying the harms they cause. ⁹⁸

IV. Toward an Alternative Design of Sex Robots

4. Sex Robots, Rape Culture, and Online and Technology-Facilitated Violence: Finding the Regulatory Connection

Debates on the regulation of sex robots are often polarized. The Campaign against Sex Robots and its supporters advocate for an outright ban, seeking to challenge the sexual commodification embodied by these machines and the socio-cultural milieu that sustains it. ⁹⁹ This policy objective is seen as so urgent and significant that it is considered to outweigh any potential benefits of their use. In this view, the fundamental right to gender equality, understood as women’s entitlement to equal rights, opportunities, and treatment across all areas of life, takes precedence over claims to sexual autonomy and its associate positive freedom to engage in sexual conduct. ¹⁰⁰ Instead, arguments for regulatory inaction emphasize respect for such autonomy and the right to privacy, arguing that restrictions on sex robots would infringe upon secular and pluralistic values by reintroducing outdated moral standards. ¹⁰¹ A less common but emerging perspective calls for targeted regulation of sex robot design, ¹⁰² particularly to address its reinforcement of gender inequalities. ¹⁰³ This approach seeks to strike a legitimate and proportionate balance between the competing interests at stake and may address the symbolic distortion of sexual consent introduced by the design and use of sex robots.

Given these tensions, recognizing the connection between sex robots, rape culture, and online and technology-facilitated violence is essential for a regulatory framework that accounts for how their design and use contribute to broader patterns of gender-based subordination and oppression. However, individuals other than women can also be negatively affected. As noted in Section 2.2, the unilateral and deceptive nature of interactions with sex robots can degrade users, while their design, primarily targeting cisgender heterosexual men, may marginalize non-cis male and non-heterosexual users. Consequently, regulatory discussions must center women’s lived experiences, expectations, and needs. This raises questions about the legal measures best suited to address the gendered harms these human-like and deceptive technologies perpetuate, reviving feminist debates on law’s role in tackling sexual violence, the limits of criminalization, and the risk of overregulation restricting sexual autonomy.

Over the past decades, much has been written about the role of criminal law in enforcing societal norms, including its use to signal that certain conduct is unacceptable, as well as the reliance on sexual consent to delineate the legal boundaries of sexual autonomy. ¹⁰⁴ However, critics argue that criminal law is inadequate for addressing gender-based violence, as its formulation, interpretation, and application have historically reinforced patterns of privilege and oppression, often to the detriment of women, non-White, and non-heterosexual communities. ¹⁰⁵ Simultaneously, the legal codification of sexual consent has faced significant criticism, especially due to its varied conceptualizations and the challenges of translating it into actionable legal evidence. ¹⁰⁶ In response, alternative measures have been proposed, particularly in the emerging field of online and technology-facilitated violence. These include civil law remedies, restorative justice, educational initiatives, and “by design” measures. ¹⁰⁷ In the European Union, evolving legislation, though not explicitly gender-sensitive and responsive, offers new possibilities. While the AI Act does not directly address sex robots, it could apply to AI-powered prototypes, ¹⁰⁸ raising questions about its role in tackling rape culture, addressing consent-distorting technologies, and responding to online and technology-facilitated violence against women.

4.1. The Transformative Potential of the AI Act for Sex Robot Regulation

Adopted in 2024 and set to come into force two years later, the AI Act aims to “improve the functioning of the internal market and promote the uptake of human-centric and trustworthy artificial intelligence (AI), while ensuring a high level of protection for health, safety, fundamental rights enshrined in the Charter, including democracy, the rule of law, and environmental protection, against the harmful effects of AI systems in the Union, all while supporting innovation” (Article 1 AIA). Although a unanimous definition of AI is lacking, Article 3(1) AIA defines AI as a “machine-based system designed to operate with varying levels of autonomy, which may exhibit adaptability after deployment. The system infers from the input it receives how to generate outputs, such as predictions, content, recommendations, or decisions, that can influence physical or virtual environments, with explicit or implicit objectives.” Consequently, a prerequisite for discussing the future regulation of sex robots under the AI Act is determining whether they qualify as AI systems under its Article 3(1).

Based on Section 2, sex robots are often equipped with the ability to learn from interactions with users, allowing for personalized and evolving responses. This capacity for adaptability creates the illusion of spontaneity and emotional connection, with sex robots adjusting their behavior based on past encounters. ¹⁰⁹ When generating outputs, sex robots process user inputs, like voice commands, and respond in ways that mimic human emotional and sexual interactions. Their responses can range from basic pre-programmed reactions to more complex behaviors that develop over time, offering a more immersive user experience. ¹¹⁰ Despite their varying operational modes, sex robots increasingly rely on AI systems to perform autonomously, with or without direct human intervention. Far from being simple machines, they are deliberately designed to influence the user’s emotional and experiential state, frequently cultivating the impression of a companion rather than just a tool for sexual pleasure. ¹¹¹

Although sex robots may fall within the scope of the AI Act, their categorization will depend on their design and functionality. The AI Act adopts a risk-based approach to regulating AI, categorizing systems by their potential risks and imposing corresponding obligations. However, as information regarding the specific types of AI systems within different sex robot prototypes is lacking, these machines could fall into any of the risk categories outlined in the AI Act, each subject to a different set of obligations based on their AI characteristics. ¹¹² Briefly, sex robots could be prohibited under Article 5 AIA if they are programmed to deceive users or exploit vulnerabilities, such as age or disability, impairing their decision-making, and causing significant harm. Alternatively, if sex robots are equipped with emotion recognition systems that infer emotions or intentions from biometric data, they would fall under the high-risk category specified in Article 6(2) AIA. This would impose strict compliance requirements, necessitating documentation and certification as outlined in Chapter III AIA. More commonly, though, sex robots are likely to be categorized as posing limited risk under Article 50 AIA, as they are designed for direct interaction with individuals, with such design being explicitly communicated to the user. Should sex robots integrate GenAI, they would require further compliance, including documentation and potentially risk evaluation obligations outlined in Chapter VI AIA. Finally, other types of sex robots may fall into the residual category of minimal-risk AI systems, which are largely unregulated unless other legal frameworks, such as the General Data Protection Regulation (EU) 2016/679, apply. In addition, the European Commission, the AI Office, and Member States may eventually introduce a code of conduct under Article 95 AIA.

With this general framework in mind, it is crucial to consider how the AI Act, despite its lack of a gender-sensitive or gender-responsive approach, ¹¹³ can inform sex robot regulation in ways that address rape culture, as well as online and technology-facilitated violence, especially in relation to programming implicit or absent consent. Such concerns reflect not only risks inherent to these technologies but also broader structural patterns of gender-based discrimination and violence embedded in society. Given that different sex robot prototypes may fall within various risk classifications under the AI Act, the creation of codes of conduct pursuant to Article 95 could play a broader role. They could provide a coherent and flexible space for consent-affirmative considerations and programming across different levels of design and production, although their voluntary and quasi-normative nature could raise concerns about legitimacy, accountability, and effectiveness. ¹¹⁴

4.2. Affirmative Consent Design: Future Directions Under Article 95 AIA

Under Article 95 AIA, the AI Office and Member States are encouraged to facilitate the creation of codes of conduct for AI systems. These frameworks foster voluntary adherence to best practices, including AI literacy, inclusivity, diversity, and safeguards for vulnerable groups, particularly regarding gender inequality. Their development may involve AI developers and a broad range of stakeholders, including civil society organisations (CSOs), academic experts, and small businesses and startups representatives. While not legally binding, codes of conduct complement the AI Act, by signaling higher compliance standards while addressing regulatory gaps. This flexibility renders them particularly relevant for AI systems like sex robots, whose design and use—as discussed above—raise questions around gender equality and sexual autonomy. These are precisely the areas where more targeted and responsive guidance may be needed, especially with regard to embedding affirmative consent into system design.

The drafting process of this code of conduct should begin with a clear recognition of the existing gender divide in the development, study, and use of technology, including AI systems and sex robots, ¹¹⁵ and must therefore involve an active effort to move beyond dominant, cisgender male and heterosexual frameworks that continue to be treated as normative defaults. Although methodologies to support this shift are still evolving, developers must begin by recognizing how their technologies may reproduce subordinating and oppressive dynamics, particularly through non-consensual interactions. They must also confront the reality that, while human sexual and emotional interactions are fluid, nuanced, and context-dependent, sex robots are often designed around simplified behavioral models that assume a uniform understanding of consent, an assumption that is fundamentally flawed. Addressing these issues should require active engagement with marginalized voices, particularly those of women. This, in turn, calls for the meaningful participation of CSOs, feminist collectives, and other relevant stakeholders, supported by adequate financial and institutional resources. ¹¹⁶ This engagement must go beyond the frequently tokenistic call to increase the number of women in science, technology, engineering, and mathematics (STEM), a framing that tends to individualize responsibility and reduce systemic exclusion to a question of access alone.

The core of the code of conduct could then center on guiding alternative, affirmative consent design approaches. Although research on consent-based design in emerging technologies is still in its early stages, a small but growing body of literature has already explored how sexual consent can be meaningfully embedded in sex robot design. This work offers important insights for how manufacturers might conceptualize consent within their codes of conduct. Notably, Anco Peeters and Pim Haselager propose equipping sex robots with a dedicated module for consent interactions. This would involve structured, reciprocal communication between the machine and its user, ensuring explicit negotiation of engagement. Crucially, the sex robot could withdraw consent and halt interaction, challenging one-sided behaviors and mitigating objectification. ¹¹⁷ Alternatively, Ellen M. Kaufman examines implementing affirmative consent, noting that some sex robots already exhibit resistance to unwanted physical contact. She suggests that gamification techniques could reinforce consent dynamics, while consent apps—despite their limited success in human relationships—could be repurposed to educate users about consent, particularly given that some machines can be programmed remotely. ¹¹⁸ Lastly, Yolande Strengers et al. advocate for visual and verbal cues, such as traffic light systems and safe words, to create interactive consent training experiences. This would involve programming sex robots to initiate and maintain consent dialogues, clarify expectations, and enforce boundaries, terminating the interaction if violated. ¹¹⁹ None of these techno-imaginaries, however, offers concrete guidance on how to operationalize affirmative consent within the design of sex robots. Yet, such operationalization is expected to occur at specific stages of the development process, sometimes in alignment with the design phases that the AI Act regulates for certain AI systems, and at other times extending beyond its scope.

This is particularly evident in dataset management, as articulated in Article 10 AIA, which addresses how data should be collected, curated, and assessed for biases and gaps that may undermine the goal of embedding affirmative consent. Research increasingly demonstrates that AI systems tend to reinforce societal hierarchies, as the data they rely on are neither neutral nor objective. ¹²⁰ Decisions through the data lifecycle, including what information to incorporate and how to classify it, reproduce these hierarchies, often misrepresenting or excluding women and other marginalized groups. ¹²¹ While little is known about the datasets informing sex robots, they likely encode gender-binary and heteronormative conceptions of sexual relations. This inference is supported not only by their existing design and usage but also by documented sexism and gender bias in AI-driven companion chatbots and human-like robots. ¹²² Accordingly, any code of conduct should include concrete measures to mitigate these risks. Based on the data feminism principles developed by Lauren Klein and Catherine D’Ignazio, ¹²³ the code should require the inclusion of diverse and representative data sources, ensure that the annotation of training data reflects a broad spectrum of consent expressions, and identify gaps or exclusions that risk perpetuating harmful norms. Furthermore, it should mandate periodic reviews of consent-related technologies, including past failures of consent apps, ¹²⁴ to understand and rectify shortcomings arising from biased models or socio-technical limitations.

While a detailed discussion of aspects, such as impact assessments, user interface design, or AI literacy, falls outside the scope of this paper, it is worth concluding by referring to the potential of recent advances in AI to support dynamic, language-based interactions. ¹²⁵ These capabilities increasingly enable sex robots to interpret and respond to natural language input, including clarifying or refining responses where necessary. Accordingly, it appears to be no inherent technical barrier to designing interactions in which users are encouraged to seek explicit confirmation that their advances are welcome, and a code of conduct could contribute to steering design choices in this direction.

V. Conclusion

This paper has examined how sex robots challenge legal and normative understandings of “rape” and “sexual consent,” situating them within broader debates on gender-based subordination and AI governance. As human-like machines designed for emotional and sexual interaction, their stereotypically gendered and deceptive nature raises concerns about reinforcing subordination and oppression, particularly when programmed to refuse sexual advances or implicitly accept them. Rather than treating this harmful design akin to certain pornography, video games, sex toys, and metaverse interactions, this paper frames it within rape culture and as a form of online and technology-facilitated violence against women.

In this context, sexual consent in user-sex robot interactions is considered symbolic. Since users often perceive sex robots as human-like, their responses and users’ reactions shape understandings of consent in gendered relationships. This paper rejects outright prohibitionist approaches and the current wave of criminalization targeting online and technology-facilitated violence against women that, especially when introduced in isolation and without broader efforts to foster a gender-equal and sex-positive socio-cultural environment, often fails to address structural inequalities, reinforces punitive logics, and risks further marginalizing affected communities rather than empowering them. Instead, it explores how the creation of a code of conduct pursuant to Article 95 AIA could promote sex robots incorporating affirmative consent mechanisms.

Ultimately, while this paper outlines a regulatory path based on a revised normative understanding of rape and consent in the context of sex robots, regulation must go beyond legal compliance. A more inclusive, participatory approach is needed, one that addresses the socio-cultural forces shaping sex robot design and use. By aligning AI advancements, particularly in sex robots, with values of autonomy, equality, and respect for fundamental rights, regulation can prevent the reinforcement of harmful gendered hierarchies and instead foster self-empowerment and expression in sexual and intimate spheres.