Who Is "the Child" in AI Law? A Feminist Perspective on Girls’ Digital Rights

An Old Category Facing a New Technology

Over the last few years, the law has developed a great deal of new language for an old concern: the protection of children online. The UN Committee on the Rights of the Child’s General Comment No. 25, the European Union’s Artificial Intelligence Act, and New Zealand’s new Online Safety (Minimum Age and Child Safety Risk Assessment) Bill each address how best to protect "children" from algorithmic and AI-related harm.^[1] However, the category performing much of this legal work, "the child", is rarely interrogated in its own right. It is treated as self-evidently universal, as though a single protective standard could represent every child’s interests irrespective of gender.

This article examines a narrower question: to what extent does treating "the child" as a gender-neutral legal category in AI and online safety regulation obscure girls’ specific digital experiences? Drawing on feminist legal theory’s long-standing critique of legal "neutrality", and using New Zealand’s new Online Safety Bill as a case study, this article argues that the law’s universal child is not, in effect, gender-neutral, notwithstanding its neutral appearance on the face of the text. The feminist response to this should not be limited to calling for greater protection for girls; rather, it should take the form of regulation that treats protection, autonomy and participation as interdependent considerations, rather than as competing priorities.

The article proceeds in five stages. It first outlines how children’s rights law currently constructs "the child" as a legal category. It then applies feminist critiques of neutrality to that construction. It identifies what becomes invisible when "children" is treated as synonymous with "girls". It subsequently tests these propositions against New Zealand’s new Bill. Finally, it proposes what a feminist approach to AI regulation might look like in practice.

"The Child" as a Legal Subject

The principal instrument of children’s rights law, the UN Convention on the Rights of the Child (UNCRC), constructs the child through a set of interlocking principles: non-discrimination (Article 2), the best interests of the child (Article 3), the child’s evolving capacities (Article 5), the right to be heard (Article 12), freedom of expression (Article 13), privacy (Article 16), and access to information (Article 17).^[2] Even on its own terms, this is a more complex subject than "vulnerable minor". The child is at once a person requiring protection and a person entitled to a voice in decisions that affect them.

This complexity was given further expression in 2021, when the Committee on the Rights of the Child adopted General Comment No. 25 on children’s rights in relation to the digital environment.^[3] General Comment No. 25 is significant because it explicitly extends the Convention to artificial intelligence, robotics, automated systems, algorithms and data analysis, confirming that children’s rights law already provides a conceptual framework through which to consider AI, rather than requiring one to be constructed anew.^[4] Importantly, General Comment No. 25 does not treat the digital environment solely as a danger from which children must be excluded. It affirms that children retain rights within that environment, including the right to participate, to express themselves and to access information, alongside the right to protection from harm.^[5] The child who emerges from this framework is accordingly not merely an object of protection. She is a rights holder whose protection needs and autonomy needs must both be given effect, rather than one being treated as subordinate to the other.

Is "the Child" Really Gender-Neutral?

Feminist legal theory has, for several decades, demonstrated that formally neutral legal categories are not always neutral in their substantive effect. Catharine MacKinnon’s influential critique of liberal legalism contends that the law’s claim to treat "everyone the same" frequently encodes the standpoint of the already-dominant group as though it were universal, such that the disadvantage experienced by other groups is never named, still less addressed.^[6] Kimberlé Crenshaw’s theory of intersectionality extends this critique by demonstrating that categories organised around a single axis of identity, whether race, sex, or, by extension, age, systematically fail to capture the compounded experience of those who occupy more than one such position simultaneously.^[7] Miranda Fricker’s account of hermeneutical injustice adds a further dimension: where the concepts available within a shared framework are shaped predominantly by the experiences of a dominant group, those whose experience falls outside that framework may struggle even to articulate what has occurred, because the language required to do so is not yet available to them.^[8]

Meredith Harbach argues that children’s rights law constitutes a natural, if underexplored, site for this form of feminist analysis, given that debates concerning formal versus substantive equality, the public/private divide, and intersectionality bear directly upon how the law conceives of children.^[9] Her argument is instructive: "the child" functions within law in a manner comparable to "the reasonable person" in other areas of law, a supposedly neutral, universal subject that was, in practice, constructed around a particular, unstated norm.

If the central question posed by feminist legal theory is whose experience is rendered visible by an apparently neutral category, the corresponding question for AI and online safety law is as follows: when "the child" is deployed as the unit of protection, what becomes of the girl situated within that category? Does she recede into an assumed, uniform childhood, or does the framework retain the capacity to recognise her specifically?

From "Children" to "Girls": What Becomes Invisible?

Three examples illustrate what is lost in the translation from "children" to "girls" within current AI governance debates.

First, it is already well established, outside children’s rights law, that algorithmic discrimination is gendered. Anastasia Karagianni’s 2025 feminist reading of the EU AI Act demonstrates that even a comprehensive, rights-based AI statute may fail to address the structural and intersectional dimensions of gender bias in AI systems, notwithstanding its formal prohibition of discrimination.^[10] That literature, however, has been developed principally with adult subjects in view. What has not yet been undertaken, and what this article seeks to contribute, is an inquiry into what occurs when this gender-and-AI analysis is applied specifically within children’s rights and online safety law, where "the child", rather than "the woman", constitutes the operative legal category.

Second, girls’ digital experience raises distinct questions concerning privacy, image and control over self-representation, questions that existing protective frameworks tend to subsume within a generic conception of "child safety" rather than treat as gendered in character. Article 16 of the UNCRC and the privacy provisions of General Comment No. 25 provide the legal basis for such analysis, yet neither instrument disaggregates by gender on its face, with the result that the gendered pattern of harm must be supplied by external research rather than by the legal text itself.^[11]

Third, and most significantly, the law tends to position young people, and girls in particular, as subjects to be protected rather than as rights holders whose views ought to shape the protective regime itself. This is not a peripheral omission; it is the proposition to which the remainder of this article returns.

The New Zealand Case Study: Protection in Practice

New Zealand’s Online Safety (Minimum Age and Child Safety Risk Assessment) Bill, introduced to Parliament on 24 August 2026, offers a timely and concrete text against which to test these propositions.^[12] The Bill would prohibit persons under sixteen from holding accounts on specified high-risk social media platforms and on AI "companion" services that simulate social or emotional connection, and would require platform operators to conduct annual child safety risk assessments addressing harms arising from recommender systems, algorithms, profiling and AI-generated content.^[13] The Bill follows a 2026 Education and Workforce Select Committee inquiry, which found the harm experienced by young New Zealanders on online platforms to be severe, and situates New Zealand within a wider international trend that includes Australia’s social media restrictions.^[14]

The purpose of examining the Bill is not to adjudicate whether it constitutes good or bad policy in the abstract, but to consider how it constructs "the child" that it seeks to protect. Its risk assessment obligations are framed in universal terms, "children", "young New Zealanders", "users under 16", with age operating as the sole differentiating variable.^[15] Gender does not feature as a category of analysis anywhere within the Bill’s stated architecture, notwithstanding that the harms it targets, including the algorithmic amplification of appearance-based content, attachment to AI companions, and image-based abuse, are well documented to affect boys and girls differently. A feminist reading of the Bill does not require a finding that its drafters erred. It requires, rather, that policymakers ask a further question that a purely age-based framework does not prompt: are the risks assessed under this Bill distributed evenly across genders, and does a uniform age threshold capture what is, in fact, occurring to girls specifically?

The Protection Paradox: Protecting Girls Without Silencing Them

Children are genuinely at risk online, and states bear real obligations to respond to that risk. However, children are also rights holders possessed of evolving capacities, entitled under Article 12 of the UNCRC to have their views heard and accorded due weight in matters that affect them.^[16] These two propositions do not align automatically. A protective measure, whether an account ban, an age gate, or a content restriction, may simultaneously operate to restrict expression, association, privacy, access to information and participation.

For girls in particular, this generates a genuine tension. Where the policy response to gendered digital risk consists simply of further restricting girls’ access, the result may be a narrowing of girls’ autonomy and voice, justified in the very name of their protection. This reproduces, in digital form, a long-standing pattern whereby protective attitudes toward women and girls have doubled as a mechanism of control. A feminist legal analysis cannot, therefore, rest at the proposition that girls face greater risk and accordingly require greater protection. It must instead confront a more difficult question: how can the law recognise unequal, gendered risk while preserving the agency of those it seeks to protect?

What Would Feminist AI Regulation Look Like?

Four principles follow from the foregoing analysis.

Gender-conscious. Risk assessments required of platforms should explicitly consider whether algorithmic and AI-related risks are distributed differently across genders, rather than assuming a uniform risk profile for "children" as a class.

Intersectional. "Girls" do not constitute a homogeneous class. Gender interacts with race, disability, socioeconomic status, sexuality and culture to produce differentiated risk profiles, and assessment frameworks should be designed to identify this variation rather than to average it away.^[17]

Rights-preserving. Protective measures should be evaluated not solely by reference to the harm they prevent, but also by reference to their effects upon privacy, expression, access to information and participation, such that protection and autonomy are weighed together rather than one simply displacing the other.

Participatory. Children, including girls from diverse backgrounds, should be afforded a genuine, structured opportunity to contribute to the rules governing their digital environment, consistent with Article 12 and with General Comment No. 25’s emphasis upon children’s participation rights in digital policymaking.^[18] Participation of this kind requires more than token consultation; it requires a demonstrable connection between what children express and what regulators ultimately decide.

Conclusion

Who, then, is "the child" in AI law? The child constructed by law cannot simply be assumed to be a neutral subject whose experience is identical irrespective of gender. Recognising this difference does not, however, require the abandonment of the universal protections afforded by children’s rights law. It requires, rather, that those protections be rendered more substantively equal, by asking whose experience the category in fact captures, whose experience it renders invisible, and who is afforded the opportunity to participate in defining what protection should mean. As New Zealand and other jurisdictions develop the next generation of AI and online safety law, the measure of a genuinely rights-respecting framework will not be the comprehensiveness with which it protects "children" in the abstract. It will be whether girls, in their specificity, are required to recede into that category in order for the law to recognise them at all.

^[1]UN Committee on the Rights of the Child, 'General comment No. 25 (2021) on children's rights in relation to the digital environment' (2 March 2021) UN Doc CRC/C/GC/25. <https://tbinternet.ohchr.org/_layouts/15/treatybodyexternal/Download.aspx?symbolno=CRC/C/GC/25&Lang=en> accessed 18 September 2026; Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) [2024] OJ L 2024/1689 <https://eur-lex.europa.eu/eli/reg/2024/1689/oj> accessed 18 September 2026; Online Safety (Minimum Age and Child Safety Risk Assessment) Bill 2026 (NZ).

^[2]Convention on the Rights of the Child (adopted 20 November 1989, entered into force 2 September 1990) 1577 UNTS 3, arts 2, 3, 5, 12, 13, 16, 17 <https://www.ohchr.org/en/instruments-mechanisms/instruments/convention-rights-child> accessed 18 September 2026.

^[3]General Comment No. 25 (n 1).

^[4]ibid (on the scope of 'the digital environment').

^[5]ibid (on children's participation and expression rights in digital spaces).

^[6]Catharine A MacKinnon, Toward a Feminist Theory of the State (Harvard University Press 1989).

^[7]Kimberlé Crenshaw, 'Demarginalizing the Intersection of Race and Sex: A Black Feminist Critique of Antidiscrimination Doctrine, Feminist Theory and Antiracist Politics' [1989] University of Chicago Legal Forum 139 <https://chicagounbound.uchicago.edu/uclf/vol1989/iss1/8> accessed 18 September 2026.

^[8]Miranda Fricker, Epistemic Injustice: Power and the Ethics of Knowing (Oxford University Press 2007).

^[9]Meredith Johnson Harbach, 'Feminist Legal Theory and Children's Rights' in Jonathan Todres and Shani M King (eds), The Oxford Handbook of Children's Rights Law (Oxford University Press 2025) ch 8 <https://academic.oup.com/edited-volume/34225> accessed 18 September 2026.

^[10]Anastasia Karagianni, 'Gender in a Stereo-(Gender)-Typical EU AI Law: A Feminist Reading of the AI Act' (2025) 1 Cambridge Forum on AI: Law and Governance e25 <https://resolve.cambridge.org/core/journals/cambridge-forum-on-ai-law-and-governance/article/gender-in-a-stereogendertypical-eu-ai-law-a-feminist-reading-of-the-ai-act/E9DEFC1E114CBE577D737EC616610921> accessed 18 September 2026.

^[11]Convention on the Rights of the Child (n 2) art 16; General Comment No. 25 (n 1).

^[12]Online Safety (Minimum Age and Child Safety Risk Assessment) Bill 2026 (NZ), Bill No 339-1.

^[13]ibid; Richard Massey and Cushla Hemingway, 'Online Safety Legislation Arrives in New Zealand: What Platform Operators Need to Know' (Bell Gully, 25 August 2026) <https://www.bellgully.com/insights/online-safety-legislation-arrives-in-new-zealand-what-platform-operators-need-to-know/> accessed 18 September 2026.

^[14]Education and Workforce Committee (New Zealand), Report on the Inquiry into Online Harms Experienced by Young New Zealanders (March 2026), as referenced in Massey and Hemingway (n 13).

^[15]Online Safety (Minimum Age and Child Safety Risk Assessment) Bill 2026 (NZ) (n 12).

^[16]Convention on the Rights of the Child (n 2) art 12.

^[17]Crenshaw (n 7).

^[18]General Comment No. 25 (n 1).