Corps de l’article

Introduction

While one might have thought that the right to abortion was solidly recognized in Western democracies, the recent situation in the United States proves its fragility. The Dobbs v. Jackson Women’s Health Organization decision, issued by Supreme Court of the United States in June 2022,[1] overturned both the Roe v. Wade (1973)[2] and the Planned Parenthood v. Casey (1992)[3] rulings on the grounds that the U.S. Constitution does not refer to abortion, notwithstanding the Fourteenth Amendment. As a result, individual states have now been at liberty to criminalize abortion in the early stages of pregnancy, provided they include an exception to safeguard the mother’s health.[4] The decision has exacerbated deep divisions between states that are not new but constitute more than a simple step backwards.

As the exercise of civil rights occurs, in this day and age, under technological constraints and is, as such, subject to digital laws and policies, artificial intelligence (AI) can be used to monitor the privacy of individuals seeking abortions and exert strong control over their bodies. Several examples illustrate its broad usage. Notably, machine learning is used to aggregate and analyze reproductive health data from multiple sources to accurately profile women (data analytics) in the context of surveillance (advanced tracking methods). AI is also central to search engines like Google Search, which actively monitor women’s online searches. On this basis, predictive AI can anticipate a woman’s intention to have an abortion and increase surveillance. It is further employed to locate and identify women approaching abortion clinics. Simultaneously, it also plays a role in disseminating information online. Recommendation algorithms integrated with AI contribute to the propagation of misinformation, exacerbating the issue of harmful content in the realm of content moderation by online platforms. AI, then, embodies a divisive power of opinions in violation of any concept of social justice. It is, consequently, imperative to tackle the surveillance capabilities of these tools as a method of subjugating women and controlling their bodies.

This two-part contribution adopts a comparative perspective, sequentially examining the U.S. and European legal frameworks. Part I delves into the online information market related to abortion, highlighting the under-regulation that allows disinformation/misinformation to flourish and impedes access to abortion rights. Part II considers the excessive technological surveillance of women and violation of their privacy when exercising their right to abortion.[5]

I. Online Information: Barriers to Access and Safeguards for Abortion Rights

The issue of online information is a vast and complex topic deeply rooted in culture. The availability of online information raises fundamental questions for our democracies and the exercise of our rights: how do we form opinions, listen to others, and ultimately build a society in an age dominated by information silos, disinformation, and online manipulation?[6] The issues become even more pressing as the use of artificial intelligence corroborates and intensifies these phenomena. The exercise of reproductive rights is no exception, with the use of online navigation as a primary method for access to medical, legal, and financial information. A multitude of questions may be posed: What risks do abortion procedures entail? Where can one find answers and advice from a qualified practitioner? Under what conditions are abortions performed? What forms of assistance are available? The dissemination of erroneous information on online platforms[7] may impact the responses to these pertinent inquiries. It should, however, be clarified that our intention is not to diminish the political dimension of this matter: individuals are free to formulate their own opinions on this deeply intimate subject. In this context, our focus is specifically directed towards the propagation of scientifically inaccurate or deceptive information concerning women’s reproductive health.

Such disinformation or misinformation threatens reproductive rights and could become a way of oppressing women and their bodies. Whether they are intentionally false (disinformation) or inadvertent inaccuracies (misinformation), such information falls into one of four categories. The first concerns false information about medication and, specifically, “reversal” procedures, which supposedly interrupt ongoing medication-induced abortions. Despite their active promotion by anti-abortion movements on social media, these procedures are deemed dangerous by U.S. authorities[8] due to the hemorrhagic risk they pose to women.[9] The second category includes the promotion of alternative medicines, for example, so-called “emmenagogue” herbs are consumed as tea and are claimed to induce a “natural” abortion. Not only is there no scientific evidence of these herbs’ efficacy, but their consumption also endangers women’s health due to their toxicity.[10] The third category pertains to the risks a woman might face following an abortion. Particularly widespread in Europe, this broad category covers physical and psychological risks. Frequently disseminated scientific inaccuracies include assertions that abortion may result in dementia, premature birth, breast cancer, and infertility.[11] Lastly, the fourth category refers to deceptive information concerning healthcare providers. This misleading information suggests that a medical centre, clinic, or professional provides abortion services when it is, in fact, a “crisis pregnancy centre” — a fake clinic opposed to abortion that aims to deter individuals from seeking out this procedure.

Digital technology thus inherently poses an early and key obstacle to the exercise of reproductive rights by compromising access to scientifically accurate information. Disinformation and misinformation simultaneously threaten access to procedures and compromise the informed consent process, as well as the ability to make fundamentally personal, health-informed choices. This section is a call for understanding and recognition of the threats to women’s ability to exercise their reproductive rights. The first part of this paper examines how digital law addresses the obstacle of digital technology. The analysis reveals that the safeguards against deceptive online reproductive information, notwithstanding their diversity, are comparatively limited (A). Nevertheless, they converge on a pivotal, albeit contentious, aspect: the significant role they attribute to online platforms in content moderation. The second part examines the role of online platforms in hindering access to abortion-related information (B).

A. Limited Safeguards of Digital Law

Addressing abortion-related information is a sensitive matter, inevitably entwined with individual moral, religious, political, and ethical convictions. The informational guarantees provided to women stand on the fine line between the right to personal choices in health and scientifically substantiated information. At the international level, the World Health Organization’s guidelines prescribe the provision of two types of abortion-related information: general and specific information. The general information is intended for the public (accurate, unbiased, and evidence-based information on sexual and reproductive health, abortion services locations, cost of services, local regulations).[12] The specific information should be tailored to each individual seeking an abortion (information required for informed and voluntary consent, potential side effects and pregnancy symptoms, and post-procedural care details).[13] These recommendations receive different acceptance levels: they face challenges regarding freedom of expression in the United States, and they are vulnerable to changes in political national governments in Europe.

The informational battle is particularly fierce in the U.S. It has long been waged in the courts, leading to surprising reversals: if the First Amendment enshrining freedom of expression was initially the foundation of women’s protection, it is now primarily mobilized by anti-abortion activists to impose significant restrictions on access to information.[14] The NIFLA v. Becerra case, decided by the Supreme Court in 2018, established a decisive milestone in this regard. The case focused on the California FACT Act that required anti-abortion pregnancy centres with a medical licence and funded by public money to inform women about the financial assistance they could receive for abortions.[15] The law sought to combat deceptive practices of anti-abortion centres, which took on the appearance of family planning centres and clinics providing abortion.[16] These deceptive practices initially occurred in person; however, they later proliferated online through the imitation of abortion clinic websites, the optimization of online searches, and the targeting of advertisement through the massive collection of personal data. Many of these centres are listed on Google as abortion clinics. The issue before the Supreme Court centred on the classification of the information provided by the centres: commercial speech (advertisements) likely to deceive consumers (according to abortion defenders), or opinion speech under freedom of expression (according to anti-abortion activists). Following a narrow and uncommon interpretation of the Zauderer test,[17] the majority opinion held that the California law violated the First Amendment by compelling centres to alter their content. It is significant that the court did not consider arguments related to women’s health, and the “professional speech” qualification from prior cases (information obligation for doctors) was rejected.[18] This decision paved the way for the anti-abortion lobby’s strategies.[19]

Defenders of abortion are also fighting back. A primary line of defence is to call on the Federal Trade Commission (FTC) to take action based on section 5 of the FTC Act that enables combat against deceptive commercial practices. In this regard, President Joe Biden’s Executive Order of July 8, 2022, encourages the authority to “consider options” concerning abortion misinformation.[20] A bill introduced in the U.S. Senate aims to grant express sanctioning power to the FTC against the deceptive practices of crisis pregnancy centres.[21] At the state level, a joint statement by California, Oregon, and Washington[22] commits to combating misinformation on abortion and implementing various measures to try to counteract the deceptive practices of crisis pregnancy centres. Connecticut’s law allows the state Attorney General to impose civil sanctions on centres engaging in deceptive reproductive health marketing.[23] An ordinance of the Los Angeles City Council allows misled individuals to bring liability lawsuits.[24]

No dedicated legislation has been enacted to combat disinformation and misinformation related to reproductive health and abortion.[25] Consequently, the content of the information is unrestricted. The main provisions available to counter abortion misinformation come from broader regulations such as the European Digital Services Act (DSA).[26] It is essential to understand that this regulation primarily points towards a method: under article 9 of the DSA, national judicial or administrative authorities can order online platforms to “act against illegal content.” Thus, the illicit content subject to removal measures is not determined a priori, in itself, but falls within the scope of legality. The informational scope, therefore, depends mainly on the policy of each state. To date, no national provision explicitly recognizes abortion disinformation/misinformation as illegal. On the contrary, attempts to limit or distort scientifically established content have flourished. Notably, Ireland previously tried to ban all communication on abortion, which was illegal in the country until 2018.[27] Other states linked information to unverified cancer risks.[28] Although the European Court of Human Rights (ECHR) recognizes the right to abortion information — which is binding on States[29] — current digital law is unlikely to have much effect on supporting this right. In the short term, platforms subjected to legal injunctions will likely remove the contentious content, but longer-term strategies are lacking.

Following the Dobbs decision in the United States, legal risk aversion has led platforms to perform a widespread removal of general abortion content.[30] To justify its removal, abortion content is often identified as sensitive and violent,[31] regardless of the nuanced local policies that divide the American territory.[32] The result is a phenomenon that has surpassed moderation both in content and geographically. This response denies women access to reliable and high-quality information about abortion, in violation of an informed choice.

This point leads to a broader exploration of the regulatory role and position acquired by online platforms regarding the phenomenon of misinformation.

B. Platforms as Gatekeepers of the Abortive Informational Market

Regarding content moderation, the United States and Europe have developed two distinct regulatory models: the United States advocates for the self-regulation of platforms, while Europe favours co-regulation.

Under American law, online platforms cannot be held liable or prosecuted for the publications of their users. This is the essence of section 230 of the 1996 Communications Decency Act, which has been designed as a fundamental guarantee of freedom of speech.[33] As a result, online platforms are free to implement their content moderation policies on reproductive health. The Gonzalez v. Google case, which questioned the interpretation of section 230, provided an opportunity to reinforce this regulatory model, particularly among advocates for abortion rights.[34] They argued that, in the absence of section 230’s shield, most platforms would be exposed to lawsuits for allowing access to information in states, like Texas, which restrict the right to abortion. They suggested that websites and online platforms could face legal action for promoting content that aids Texas residents in obtaining an abortion, in violation of SB 8.[35] As a result, according to them, without section 230, most websites would choose to limit the legal risks by removing any abortion-related content. In a joint letter, they warned the Attorney General that “online services might be compelled to limit access to reproductive resources, for fear of violating various state anti-abortion laws.”[36] Nevertheless, a lack of liability does not mean platforms should evade responsibility. As the final section describes, this is a flaw in the model: platforms do not ensure access to scientific information on reproductive health. Before delving into the platforms’ policies on this matter, let’s briefly explore the European model.

Conversely, European law obligates platforms to ensure a moderation policy in line with the rule of law and subject to penalties. Three legislative measures are particularly relevant to guide their policy on reproductive health. First, article 26 of the DSA requires online platform providers to ensure that “for each specific advertisement presented to each individual recipient”, the recipients can “in a clear, concise and unambiguous manner and in real time” identify the natural or legal person on whose behalf the advertisement is presented (b) or who funded it (c). The text adopts a broad definition of advertising.[37] This transparency requirement[38] should help identify health providers while reducing the deceptive effect of sites mimicking governmental policies. Article 26 of the DSA is strengthened by the EU’s Code of Practice on Disinformation, which binds its signatories under article 45.[39] This article provides that online platforms commit to “put in place or further bolster policies to address both misinformation and disinformation across their services, and to agree on a cross-service understanding of manipulative behaviours, actors and practices not permitted on their services”.[40] Second, the DSA requires platforms to track professionals: by collecting their information and assessing their reliability when the intermediation service provided by the platform allows professionals to offer their products or services to EU consumers.[41] This measure aims to counter inaccurate web referencing, which leads to false clinics. Third, under article 34, very large online platforms are subject to an audit aiming to identify the “systemic risks” raised by their recommendation systems, content moderation systems or general terms and conditions, as well as their implementation.[42] Among the identified systemic risks are “any actual or foreseeable negative effects for the exercise of fundamental rights, in particular the fundamental rights to human dignity”, and “any actual or foreseeable negative effects in relation to gender-based violence, the protection of public health and minors and serious negative consequences to the person’s physical and mental well-being”.[43] These risks must be mitigated to avoid sanctions. The regulation explicitly targets the fight against disinformation campaigns. This tool could encourage online platforms to prevent the widespread dissemination of misleading information and ultimately protect a woman’s ability to make informed decisions about her health.

Given the recent entry into force of the DSA, it is premature to assess the effectiveness of its tools in ensuring and protecting reproductive rights. However, both the US and European regulatory frameworks result in the same fact: whether mandated or voluntary — and differently exposed to the state constraints mentioned — content moderation relies on private operators, namely the platforms. These observations call for an examination of the policies implemented through their terms of use, on the one hand, and their practices, on the other.

Despite some announcements advocating for the protection of women’s reproductive health following the Dobbs decision,[44] online platforms remain largely silent about the guarantees provided. Only YouTube and TikTok mention it in their terms of use: the first prohibits content contradicting advice from local health authorities or the World Health Organization concerning the safety of medical and surgical abortion methods;[45] the second prohibits advertising abortion services and the diffusion of misinformation in the American market.[46] X, formerly Twitter, restricts the promotion of health and pharmaceutical products and services. Terms and conditions of use require prior authorization for advertising (notably) abortion clinics and advocacy.[47]

Regarding practices, empirical studies show that the spread of false information about abortion has significantly increased in the last two years. In the United States, 83% of Google searches regarding abortion refer to reversibility procedures,[48] which the search engine considers to be “safe and effective” techniques.[49] Meanwhile, YouTube has failed to remove videos promoting false information. Despite its terms of use and its promise to label and identify content approved by scientific studies, the platforms initiatives are marginal: they focus exclusively on English-language videos.[50]

It also appears that the online platforms’ monetary interests influence moderation. Dissemination of misleading and/or false content on abortion appears to be particularly lucrative. Targeted advertising about reversibility procedures generated significant profits in the United States.[51] The substantial funding of anti-abortion movements also ensures the consistent online presence of these narratives. These figures derive mainly from North America, as no equivalent empirical European studies were found. However, experience in content regulation indicates that policies are only partially regionalized.[52] Observational data also suggests that the regulation by platforms is susceptible to private interests, influencing moderation policies and, consequently, guarantees in terms of reproductive health. The dominance of corporate economic interests over the human rights of women must be challenged and actively resisted.

We have described how the regulation of the online informational market is threatened by both public capture (censorship for public policy reasons) and private capture (exposure to conflicts of interest). The digital realm — and its law, to some extent — thus contributes to hindering access to the abortion right. The second part of this article focuses on the exercise of this very right. Digital means then become repressive tools of pro-natalist policies, in disregard of reproductive health rights. Once again, such practices must be tackled to preserve women’s rights.

II. Digital Enforcement of Policies Penalizing Abortion in the Post-Dobbs Era

Following the Dobbs decision, means of digital surveillance were mobilized by states that had adopted abortion criminalization laws (Alabama,[53] Arkansas,[54] South Dakota,[55] Oklahoma,[56] Louisiana[57]) to ensure their enforcement.[58] In contrast, progressive states (Washington,[59] California,[60] Massachusetts[61]) enacted protective laws to limit this society of surveillance (A). Even if, in the European Union, the use of technology is not as oppressive as in the United States, it is equally urgent to assess the ability of digital laws to protect reproductive rights in the age of AI, threatening respect for human rights (B).

A. Digital Surveillance of Reproductive Rights in the United States

The use of digital surveillance in criminal matters is not new. In this context, however, it supports pro-natalist policies and particularly affects the privacy and exercise of women’s reproductive rights. A report published in May 2022 by the Surveillance Technology Oversight Project (STOP)[62] highlights that conservative state legislators are pressuring police and prosecutors to use all tracking tools available to target pregnant individuals and healthcare providers.[63] The diversity of digital means involved is alarming: collecting search engine data;[64] recording electronic payments on retail sales of abortion pills, over-the-counter medication and prescription medication;[65] collecting mobile phone data[66] and menstruation tracking app data;[67] and monitoring electronic communications[68] (e.g. emails, social media messages, communications from video games). For example, Meta’s “Messenger” service shared conversation data between a teenager and her mother with the Nebraska police to prove an illegal abortion.[69] It is worth clarifying that in this case, Messenger did not share the conservations. It was in fact Meta that disclosed the details of the conversation that occurred via Messenger after receiving a copy of the police warrant concerning the exchange. This example is important as it highlights the complicated relationship between Meta’s obligation to comply with disclosures subject to a warrant (often explicitly mentioned in an organization’s privacy policy) and an individual’s right to privacy. This case is not isolated.[70] Access to advice and reproductive services is increasingly online,[71] exposing individuals to liability in anti-abortion states.

Digital surveillance also occurs through the collection of geolocation data from mobile phones around abortion clinics and identification data from body cameras used by anti-abortion activists, as well as automatic licence plate reading.[72] Data can also be shared by “crisis pregnancy centres,” located near family planning centres, to deter individuals from seeking abortions.[73] On another note, AI can enhance an organization’s location tracking capacities, for example, by applying computer vision technology to photos and videos.[74] Geolocation, notably, poses a risk in states such as Idaho, where cross-state movement to abort is prohibited.[75] Location data, then, constitutes potential evidence of a state crime.[76] This tracking method is also relevant to online activity. For example, Google claimed to exclude abortion clinics from users’ location history,[77] but investigations by Accountable Tech[78] and The Washington Post[79] revealed that this promise was not kept.

The collection of personal data is further facilitated by the fact that entities subject to federal laws regarding health (through the Health Insurance Portability and Accountability Act of 1996),[80] financial data (through the Gramm-Leach-Bliley or Financial Services Modernization Act of 1999),[81] and electronic communications (through the Stored Communications Act of 1986)[82] must respond to criminal investigations and, in many cases, to warrants and subpoenas as well. Moreover, digital actors are not within the scope of these laws, meaning that brokers[83] can sell data to police services, outside of any judicial and legislative control. Furthermore, criminal investigations are now characterized by general warrant requests, based on a broad spatial perimeter by geofence around abortion clinics[84] as well as by online search keywords.[85] These “digital dragnets” allow the identification of a large number of abortion seekers,[86] even if the Supreme Court of the United States has declared that the Fourth Amendment of the Constitution prohibits the use of such warrants in the absence of evidence of a “probable cause.”[87]

In response to the overcollection of personal data and in order to reinforce the respect of the women’s reproductive rights, progressive states are defending reproductive rights by enacting data protection legislation. The state of Washington was the first to enact such legislation with the My Health My Data Act in April 2023,[88] which requires tech companies (e.g. social networks apps) to obtain explicit consent before collecting and selling health data. This includes information on sexual and reproductive health collected by menstruation tracking apps, as well as location data that may indicate a consumer is receiving health services. Geofencing within a perimeter of 2,000 feet around abortion clinics is prohibited.

California also passed a series of thirteen laws in September 2022 to expand access to abortion.[89] The Reproductive Rights Act[90] prohibits law enforcement agencies, as well as California companies providing electronic communication services, from complying with requests (arrest or search warrants) from law enforcement agencies of another state[91] or federal agency[92] if abortion is legal in California. Another law, the Confidentiality of Medical Information Act: reproductive or sexual health application information (CMIA), was enacted in September 2023[93] to protect data collected by mobile apps or websites that collect reproductive or sexual health information. These services will have to comply with the same medical information confidentiality standards as traditional healthcare providers. California legislation additionally prohibits government entities from submitting, and courts from executing, a “reverse keyword” or “reverse location” request by judicial warrants.[94] In Massachusetts, the Location Shield Act[95] prohibits brokers from selling mobile phone location data to third parties, requiring law enforcement to obtain a warrant.

While the measures taken by progressive states are a step in the right direction for the defense of women, they are often limited in scope. As a result, every new protective rule granted to women is immediately denied by conservative states. Such a legislative approach is therefore not sufficient to protect women’s reproductive rights in a sustainable and effective way. Moreover, the efforts made in the U.S. to regulate AI at the federal level are very limited. The lack of legislative agreement before Congress led the Biden administration to publish an executive order.[96]

In this context, it is important to see if the European Union succeeds in implementing a legislative model that is truly favorable to women. At first glance, the exercise of the right to abortion seems less controversial in Europe than in the United States, but the potential for a setback in reproductive rights is worrisome in the era of AI.

B. Digital Surveillance of Reproductive Rights in the European Union

The European Union is built on common values and fundamental rights. However, the Charter of Fundamental Rights of the European Union[97] does not enshrine a right to reproductive freedom. Additionally, while all member states allow abortion, significant legal and practical disparities exist. The rise of conservatism in countries like Poland or Hungary comes with threats, both in law and in practice, to the exercise of abortion. In Poland, a 1993 law[98] banned abortion except in three cases and one of the exceptions was declared unconstitutional in October 2020 by the country’s Constitutional Tribunal[99]. This decision was later confirmed through an amendment of the law in January 2021.[100] Now, doctors in Poland face imprisonment for performing abortions deemed illegal. Consequently, Polish women often turn to online sources for abortion pills or travel abroad for the procedure,[101] making them vulnerable to digital tracking. This risk is heightened by the Polish government’s creation of a pregnancy declaration and tracking database,[102] purportedly to improve patient care. Yet, there are concerns that this information could be used to crack down on abortions outside the country or to prosecute healthcare professionals.[103] Thus, while digital surveillance is not as widespread as in the United States, the collection of reproductive health data exerts pressure on those seeking abortions and on medical staff.

Given the growing threats to reproductive rights and risks of technological surveillance, can European digital legislation serve as protective barriers? Besides the aforementioned DSA, the General Data Protection Regulation (GDPR)[104] protects personal data and is complemented by the AI Act.[105] Article 9(1) of the GDPR prohibits the processing of sensitive data, including health data and data on sexual life or orientation. However, article 9(2) provides exceptions that permit such processing if the individual has given explicit consent. This may apply when using menstrual tracking apps or health information websites, but also when sharing delivery information to order abortion pills online. Individuals might, furthermore, agree to location tracking through mobile apps like Google Maps when visiting healthcare facilities. Still, the legitimacy of this data processing becomes questionable in light of opaque automated data processing practices and potential information sharing with law enforcement. There are also doubts regarding compliance with principles of legality, loyalty, necessity, purpose, and minimization which stipulate that data cannot be used outside its collection purpose.[106]

Additionally, individuals have rights, such as the right to “not to be subject to decisions based exclusively on automated processing, including profiling”, that produce legal or similarly significant effects on them.[107] Yet exceptions, especially those related to the laws of Members States, diminish this right.[108] With Poland’s national registry authorized by local law, automated processing might apply to the collected data, potentially profiling women. Procedural safeguards, like the need for a judicial warrant,[109] might be insufficient.

Moreover, article 23 of the GDPR allows for limitations to individual rights when member state law provides for measures related to crime prevention and detection, investigations, prosecutions, or penal sanctions execution. Thus, digital entities could share user personal data upon state requests. These limitations must, however, respect fundamental freedoms and rights and be necessary and proportional in a democratic society. But without a reproductive freedom protection right in the EU Charter, how would the Court of Justice interpret these provisions? Would it rely on privacy protection and personal data in a context where abortion is criminalized nationally?[110] Its jurisprudence leans on the subsidiarity legal principle, deciding that “[i]t is not for the Court to substitute its assessment for that of the legislature in those Member States where the activities in question are practised legally.”[111] It merely recognizes abortion as a service, per article 57 of the Treaty on European Union,[112] allowing individuals to benefit from it in more permissive states without interfering in national laws to defend women’s reproductive rights. Consequently, such an interpretation does not recognize these rights throughout the European Union.

Furthermore, the AI Act aims to apply rules based on the risk levels of AI systems. High-risk systems prevention relies on meeting requirements for use cases listed in Annex III of the regulation.[113] However, according to the final version of the AI Act, the AI use cases outlined in this contribution are not included in the category of high-risk AI systems. Furthermore, it prohibits

the placing on the market, putting into service for this specific purpose, or use of an AI system for making risk assessments of natural persons in order to assess or predict the risk of a natural person to commit a criminal offence, based solely on the profiling of a natural person or on assessing their personality traits and characteristics.[114]

This prohibition does not apply to AI systems used “to support the human assessment of a person’s involvement in criminal activity, which is already based on objective and verifiable facts directly related to the criminal activity.”[115] It is highly likely that the various search processes used to track down women could collect “objective and verifiable facts” of such a nature as to fall within the exception. The interpretation of the text raises concerns: should the issue of digital surveillance of women seeking abortions not command greater attention within the values upheld by the EU? Women whose reproductive rights are criminally threatened in their countries are left insufficiently protected by the European Union.

Finally, law enforcement agencies are allowed to use high-risk AI systems

intended to be used by or on behalf of law enforcement authorities or by Union agencies institutions, agencies, offices or bodies in support of law enforcement authorities for profiling of natural persons as referred to in article 3(4) of Directive (EU) 2016/680 in the course of detection, investigation or prosecution of criminal offences.[116]

While high-risk systems are bound by specific obligations, such as implementing data governance measures to mitigate bias and errors, the use of an AI system for profiling individuals during the detection, investigation, or prosecution of criminal offenses is allowed. The regulation sets rules to minimize these systems’ risks without banning them outright. Considering the risks analyzed in this paper for women seeking abortions, the European Union has not adequately recognized the potential threats to reproductive rights in a post-Dobbs era dominated by American tech firms. At this stage, we can only hope that the Court of Justice will interpret the AI Act in a way that protects women.

In this contribution, we have explored the diverse implications for women’s reproductive rights within the realm of technology, specifically AI, spanning both the United States and the European Union. The first part demonstrated the pervasiveness and perils associated with online misinformation and disinformation surrounding abortion, emphasizing the heightened risks posed to women. The reluctance in the United States to adopt content moderation measures, coupled with the robust protection of freedom of expression, exposes women to risks and obstructs their access to quality information on reproductive healthcare, especially abortion. Meanwhile, in the European Union, the recently implemented measures under the DSA appear hopeful, but their real effectiveness remains uncertain. The second part showed how surveillance technologies can directly enforced criminal abortion policies, adversely affecting women’s rights. Not only the access to abortion can be compromised, but their right to privacy and dignity can also be violated. The European Union provides more protection for the right to privacy and the personal data through the GDPR, but this text is unable to prevent a State from implementing a criminal policy and authorizing the processing of personal data for this purpose. Lastly, the AI Act addresses unacceptable and high-risk AI systems, but certain applications by repressive authorities receive exemptions (unacceptable AI) or fall under the high-risk AI category, making them permissible. As a result, women are inadequately protected from the risks we have identified.

Therefore, there is an urgent need to recognize the surveillance capabilities of AI tools as instruments of women’s oppression, safeguarding their reproductive rights for enhanced social justice. In this, like other fields, the restorative power of AI still needs to find its place. In the U.S., Danielle Citron encourages to consider certain forms of intimate privacy as a civil right.[117] In the EU, privacy and data protection are already fundamental rights,[118] however there is a need to go further. While the AI Act seeks to mitigate discrimination risks, it falls short in adequately and systematically addressing gender effects in the use of AI. The exercise of reproductive rights is just one example among many, and this is crucial to advocate for a better understanding of the negative impact of AI on women and for a more robust legal framework to ensure a better gender balance. One of the solutions could be to encourage the European Court of justice to interpret the AI Act and reinforce the non-discrimination provisions, through articles 21 and 23 of the EU Charter of Fundamental Rights.[119] Article 21(1) prohibits any discrimination based on sex and article 23(1) states that “equality between men and women must be ensured in all areas.”[120] It is time to guarantee such equality within AI and digital technologies. Moreover, the principle of equality shall not prevent the maintenance or adoption of measures providing for specific advantages in favor of the under-represented sex. Feminist Science and Technology Studies have, notably, yielded substantial evidence of discrimination against women.[121] This body of knowledge must be translated to the legal and judicial framework to address the inequalities they face.