Justice requires that citizens enjoy basic security. By “basic security,” I mean effective protection from physical violence by other agents, e.g., assault, kidnapping, and murder – and not protection against disasters or other non-agential dangers or risks. A just society requires that citizens enjoy basic security of the person against threats to life and physical assault. I will argue that security of the person must be at the highest level of priority because it is a precondition of enjoying other rights and liberties.

Security is noted as a right by many liberal theorists. John Locke wrote that reason teaches us that, according to the law of nature, “no one ought to harm another in his Life, Health, Liberty, or Possessions.”6 Basic security, of course, protects “life, health, and liberty” against harm by others. John Rawls includes “integrity of the person” among the liberties in Political Liberalism and the revised edition of Theory of Justice.7 And Robert Nozick argued that the (minimal) state was justified as an association protecting security, including one’s external property as well as one’s body, “against force, theft, fraud, …”8

But, as Henry Shue influentially argued, security stands in a special relation to all other rights. Shue defines the right to security as a right “not to be subjected to murder, torture, mayhem, rape, or assault.”9 In Shue’s terminology, security is a basic right. Basic rights are those rights which are necessary for or presupposed “in both the enjoyment and the protection of all other rights. Other rights could not be enjoyed in the absence of security or subsistence, even if the other rights were somehow miraculously protected in such a situation.”10 Although we might, for example, be able to run from Point A to Point B “miraculously” unharmed in an area where violent attacks are widespread, we cannot be said to enjoy the right to freedom of movement if bullets are whizzing past in all directions: “from the viewpoint of the bearer of a right it would be false or misleading to assert that a right had been fulfilled unless in the enjoyment of the substance of that right, a person also enjoyed protection against the threats that could ordinarily be expected to prevent, or hinder to a major degree, the enjoyment of the initial right.”11

Security is thus a basic right (like subsistence, on Shue’s view) because it is a condition for the protection and enjoyment of all other rights: “No one can fully enjoy any right that is supposedly protected by society if someone can credibly threaten him or her with murder, rape, beating, etc., when he or she tries to enjoy the alleged right. Such threats to physical security are among the most serious and … widespread hindrances to the enjoyment of any right.”12 Without the right to security, one could not reliably exercise rights to freedom of movement, free speech, political participation, freedom of association, and so on.

Shue is a human rights theorist, but these points regarding the enjoyment and protection of liberties apply across liberal frameworks. If people cannot reasonably expect freedom from arbitrary violence, they cannot reliably exercise their liberties, and so security of the person is a condition of reliably exercising the liberties. Security is thus a precondition to a just society: a society in which we could not regularly and reliably go about our daily lives without assault is not one in which liberties could be reliably exercised. Exercising the liberties depends on having the right to move about and speak without assault or the constant threat of assault.13 In James Nickel’s words, security rights are “essential building blocks for a system of liberty.”14

Security rights thus must be at the highest level of priority for liberal theories of justice, because securing them is a necessary condition for securing all the liberties.15 (Even though one might “miraculously” enjoy that which the liberty protects in the absence of security, one cannot say that the liberty itself is protected if one cannot exercise it securely.) As Shue writes, because basic rights are needed to protect all other rights, “giving less priority to any basic right than to normal non-basic rights is literally impossible.”16 Any theory of justice committed to liberties must thus accord the highest priority to security in the design of its institutions. The priority of security grounds a claim that the state take effective means to protect security (subject to epistemic burdens, feasibility, and conflict with other liberties). This implies that the design of legal institutions should protect security, with at least equal priority to (other) liberties. (Security is not technically a liberty, but I refer to it thus, loosely.17)

This claim about the priority of security is not a normative claim about what justice demands, but a logical claim about the precondition for the protection of any other rights.18 Nicholas Southwood and Robert Goodin have helpfully distinguished two senses of the phrase “justice requires.” This phrase, they note, is “ambiguous between claims involving requirements-as-necessary conditions for justice and requirements-as-demands of justice.”19 That is, we can distinguish between what is required as a necessary condition for justice and what is required as a deontic demand of justice addressed to certain individuals or collectives.20 Shue’s argument, applied to liberal theories, suggests that security is required as a necessary condition of liberal justice, because liberties cannot be enjoyed in the absence of security. And this requirement (in the sense of necessary condition) gives the protection of security a priority.

To clarify, I am not arguing that security should have absolute priority; a moment’s reflection will show that not all other goods should be sacrificed to security. Security must be weighed with other liberties and moreover, it requires effective protection, not every conceivable protection. (And, as Shue argues, it only requires protection against commonplace threats, not every conceivable threat.21) For example, prohibiting and redacting all discussions of violence, including in journalism, works of history, and works of literary merit, would (in my view) clearly be going too far, even if it were demonstrated that this could somewhat reduce the incidence of violence; plausibly, this is not required for effective protection against commonplace threats. But it is clear that there may be more realistic, and more difficult, cases of conflicts between security and liberties such as freedom of speech, religion, movement, and association.

While I will return to these conflicts briefly below, how to resolve such conflicts with liberties and how much protection is required are important questions outside the scope of this article (and different theories will yield different approaches to settling this question). Dealing with tradeoffs among liberties, and among risks, raises very difficult questions.22 My point is simply the modest, and general, one that the effective protection of basic security must be at the highest level of priority.

One point about tradeoffs should be noted here, however. The distribution of security matters. Security should not be interpreted merely as a requirement to prevent the most violent assaults, no matter their distribution. Because security is required by justice as a precondition for enjoying rights, theories committed to the equal protection of liberties are committed to the equal protection of security. Thus, equal distribution of security (i.e., effective protection against violence) among different social groups matters. A policy which saved the most lives would be unjust if it saved only the lives of the wealthy while the worse-off were unprotected. To take another example, Amia Srinivasan cites a study suggesting that mandatory arrest policies in IPV increase violence against black women, through retaliatory violence, while deterring violence against white women.23 A policy which inadvertently increases violence against black women while reducing violence against white women exemplifies the distributive point: this distribution of protection is unjust. The distribution of security matters from the standpoint of justice.

Finally, protecting basic security has priority over protection against non-agential risk and hazard, such as disease and disaster, as well as over non-directed risks arising from negligence. An assurance that basic security is protected allows us to interact in relations with others in which violence is prohibited and we have some confidence that the prohibition is effective. We cannot interact with others in decision-making or cooperation if we do not feel some assurance that they will not violently assault us. This is a reason, one which links basic security to citizens’ status as equal rights-bearers, for elevating the protection of basic security over protection against other kinds of risk.

Before proceeding to ask how the liberal state should protect security, I will make one further point about taking IPV seriously as a security threat against which the state should provide effective protection. I have defined “basic security” in an uncontroversial way as protection against physical violence: (almost) everyone agrees the state should play a role in addressing physical assault. But taking IPV seriously suggests we should consider larger patterns of abusive behavior, some of which does not consist in physical violence. IPV predominantly occurs as a pattern of behavior including isolation, threats, financial control, and monitoring of behavior (termed “coercive control” by Evan Stark).24 This led Claudia Card to remark that domestic violence “is a campaign, a policy, not an isolated episode.”25 IPV is a distinctive type of violence: repetitive, patterned, not all physical.26 A focus on single acts of physical violence may obscure threats to security. When the law focuses on single incidents of violence, it is “unable to ‘see’” the repetitive nature of the violence and may consequently underestimate it.27 Likewise, when the law focuses only on physical violence, it may ignore other behaviors, such as emotional abuse, which enable and compound physical violence.28

In favor of extending “basic security” to include a larger pattern of violence which is not all physical, consider that agents’ exercise of their rights, and more generally, pursuit of their plans, requires not just that the state take effective means to protect security, but that agents have warrant to believe that they can proceed securely. Such warrant may be undermined by a persistent atmosphere of threats and control, even when physical violence is not present.

Despite these considerations in favor of expanding the definition of security beyond protection from physical violence, my argument does not rely on expanding it. The argument of this article depends only on interpreting security more ecumenically, as “effective protection from physical violence by other agents.” But the interventions I argue for may in effect protect against the larger pattern of coercive control as well.

Having set out the definition of basic security and its priority, we can move forward. It might be thought that the criminal justice system has a special place in protecting security, since punishment might deter violence, as well as removing users of violence from society, expressing social disapproval of violence, or preventing escalating retaliation.29 Rawls, for example, suggests that security justifies the penal system in a Hobbesian way, because the threat of punishment provides security through mutual deterrence.30 But, as I will argue is the case with IPV, deterrence through the penal system does not always adequately protect security.

The priority of security implies that security should figure in any legal institutions in which effective protections can be implemented; its protection should not be limited to the criminal justice system. To take a low-level example, security is just such a consideration when a council places lighting on a dark street to deter opportunistic violence. Likewise, marriage and family law may be designed to protect security by protecting against vulnerability to abuse. Limiting the protection of security to criminal justice and the penal system is arbitrary, if other institutions can protect it more effectively (without countervailing costs sufficient to block or outweigh the protection). Other legal institutions may protect security and may be needed to do so effectively. There is no principled reason to restrict protection of security to criminal law, if other institutions can protect it as, or more, effectively.

It might be thought that, unlike other liberties, security should only be upheld in a backward-looking way – that is, that the state has no special role in protecting against security risks, but it should punish security violations. But, for one thing, this would differentiate it arbitrarily from other liberties, which are protected across institutions. For another, a common argument for the penal system is that it protects security by deterring violence – for those holding such a view, its rationale is to reduce risk of violence, not (simply) to punish offenders. But punishment is not the only – and arguably not the most effective – way to protect security.

So far, I have been defending the priority of security at the theoretical level. But how would the priority of security be applied in designing institutions? The priority of security suggests that (so far as it can) the state should take effective means to preventing interpersonal violence. Surely this implies that it should take the most effective means among alternatives when there is no countervailing cost; it might also require that the state provide effective protection above a certain baseline (e.g., reducing incidences of assault to a certain rate). The priority of security also implies that where legal institutions can be structured to protect more effectively against it (without countervailing costs), the state should do so.

Institutions can seek to protect security directly or indirectly. Direct protections offer protections against vulnerabilities and interventions aimed directly at stopping violence or allowing escape, such as a safe-house to shelter the victim. Direct interventions targeting violence may be long-term, such as job counseling within the safe house to help the victim become independent or educational programs aimed at decreasing violence. Indirect approaches try to reduce the incidence of violence by addressing its precursors or social determinants. For example, Goodmark argues that because male unemployment correlates with domestic violence, reducing unemployment is an effective means of reducing IPV.31 Direct protections will be the main focus of my argument. This is not because I am ruling out indirect approaches. However, because they aim at protecting security by addressing some other thing (such as unemployment), they less clearly fall within the rationale for the priority of security which I have given. They aim not at protecting security directly, but as a side effect of some other intervention. However, with adequate empirical evidence of their effectiveness, there well may be a strong case for such indirect measures as a means of protecting security.32

Intimate relationships are security risks. They create distinctive opportunities for violence. Such opportunities are not limited to marriage or marriage-like relationships, but they also exist in friendships, between adult children and elderly parents, and between parents and children. I have argued that basic security must be a high priority for the state because insecurity constrains the exercise of liberties. Yet, for victims of intimate partner violence, such insecurity is common, limiting the exercise of their liberties to some degree. For example, a victim of IPV may be limited in her freedom to associate, her freedom of occupation, even her right to vote.

Indeed, many philosophers have argued that some vulnerabilities are intrinsic to valuable relationships and such vulnerabilities reflect or deepen their value - for example, making oneself vulnerable to another can build trust.33 But the same features of intimacy, access, and trust which can make such relationships valuable also make them sites of vulnerability to wrong-doing.34 There is a long history of feminist work on women’s vulnerability, in particular, within marriage and intimate relationships.35

I argue, drawing on some of this work, that intimate relationships create vulnerabilities to violence and that these vulnerabilities also produce distinctive barriers to enforcing criminal law against IPV. I focus on vulnerabilities to violence in adult intimate relationships. Some of these vulnerabilities may be inherent to intimate relationships and hence intractable. Some vulnerabilities are created by the state. (For example, marriage law may exacerbate some pre-existing vulnerabilities and create new ones in preventable ways.) In either case, to the extent that vulnerabilities to violence can be addressed through protective measures, they should be addressed by the state as a means of protecting against a widespread threat to security.

To be clear, by “intimacy,” I mean physical, informational, and emotional access not granted to strangers; this comes in degrees.36 It is not limited to romantic or sexual relationships, but extends to friendships and family relationships. By “relationship,” I mean an evolving series of interactions between particular persons over time.37 Intimate relationships, through their essential features of evolving patterns of access over time, create vulnerabilities to violence. Feminists have drawn a great deal of attention to one way that this happens, typically in gender-structured relationships: exit options are limited as economic dependency develops. But a second kind of vulnerability, which I call “vulnerabilities of access,” arise from the features of intimate relationships themselves.

In this section, I will address three objections: first, that effective interventions will be paternalistic and hence illiberal; second, that effective interventions will invoke a contested conception of the good relationship and hence be deemed illegitimate by liberals committed to neutrality; third, that effective interventions will conflict with the protected private sphere or protected liberties.

Concerns about state paternalism are found throughout contemporary legal literature on IPV. For example, Goodmark writes that mandatory arrest and no-drop policies deprive women “of the ability to determine whether and how the state will intervene in their relationships.”64 This suggests the objection that the strategies I propose are paternalistic, that they wrongly interfere in victims’ decision-making on the ground of what the state deems to be their good.

A first reply is that the protections I propose offer choices. Legal safeguards can be used or not, and entry into formal institutions is voluntary. These protections do not restrict but expand choice. However, the objector could pursue the point: such choice architecture interferes in choice by shaping options on paternalistic grounds. Designing marriage law on grounds of the participants’ own good appears to engage in paternalistic judgments. Likewise, offering ways to end intimate access within relationships in some cases (such as revoking access rights to a shared domicile or accounts) appears to substitute the state’s judgment of what is in the victim’s best interests for what she had previously chosen (and would otherwise be bound to). Seana Shiffrin makes such a point regarding unconscionability doctrine in contracts: “even some freedom-enhancing behavior can be paternalist,” she writes, when someone provides options for another, against that person’s expressed preferences or even before they have chosen, on grounds of that person’s best interests. For Shiffrin, paternalistic policies are those which substitute the paternalist’s judgment for that of the person affected, in an area in which the affected person has the authority to make their own decisions.65 This would seem to apply to designing marriage law, or offering protections which might override a victim’s earlier commitments.

However, Shiffrin’s defense of unconscionability doctrine provides a response to this worry.66 While a liberal state should not substitute its judgment of citizens’ best interests for their own in areas in which citizens have authority to choose, the state can refuse to recognize or enforce contracts which it find exploitative or morally objectionable. Recognizing citizens’ right to make certain decisions does not require the state to use its resources to enforce those decisions – particularly when this might amount to enabling “harmful, exploitative, or immoral” behavior. In such cases, Shiffrin writes, “metaphorically … the state refuses, for its own sake, to be a codependent.”67 Similarly, the state could justify policies such as I have described, not on the paternalist grounds that it knows better than citizens what is good for them, but on the self-regarding grounds that it refuses to enable and enforce opportunities for violence.

This (particularly the use of moral language) may prompt another concern for liberals committed to neutrality, namely that the choices offered reflect an illegitimate judgment by the state that certain relationships are undesirable or immoral. This could be framed as an objection on grounds of perfectionism, that the institutional choices offered wrongly invoke a contested conception of the good relationship. After all, the objector might urge, parties’ choosing a relationship involving consensual violence – such as BDSM – surely reflects a conception of the good with regard to which the state should be neutral.68 But this could extend to cases in which parties would consent to remain in an abusive relationship, say, on religious grounds.

Two examples will help clarify the worry.

However, such interventions embedded in public education and marriage law can be justified by a neutral rationale. Basic security, as part of the content of justice and a logical precondition for it, is a neutral rationale, which can justify intervention.

Consider an analogy with public health campaigns: the state issues warnings about certain behaviors to promote health. If these warnings were grounded in a vision of human excellence as involving athleticism, say, they would be illegitimately perfectionist. But as grounded in the primary good of health, these warnings do not (on that count) violate the constraints of neutrality. Just so, warning about the security risks of relationships, or intervening in relationships to protect security, does not commit the state to a comprehensive conception of the good relationship. Protective state intervention against vulnerabilities arising in relationships can be justified to the neutral liberal on grounds of security.

A third, and perhaps the most difficult, problem is that effective interventions risk violating privacy. The history of IPV and privacy is thorny. Historically, the private sphere, construed as marriage and family, was shielded from justice.71 Echoes of this view can still be found in late 20^th century philosophical writing on marriage, e.g., the idea that the valuable private sphere must be protected from external interference.72 But an assault in an intimate relationship or a marriage should not be mitigated just because it occurs in an intimate relationship or a marriage. The private sphere should not be shielded from justice as such. Historical justifications for the exclusion of the private sphere from justice – such as that married couples were one person and so could not wrong each other, or that affection would regulate families so there was no need for justice – face serious problems.73

Moreover, the notion of marriage and the family as a “private sphere” needs rethinking. The state intervenes in relationships in many ways, in structuring their terms (in marriage and family law) and flagging them as eligible for or liable to certain provisions (tax law, healthcare, zoning, employment, residency, housing policy). Law constructs the supposedly private sphere of intimacy and family and undergirds social norms and practices. It presents one family path as normative, with legal and social penalties for those who deviate: examples include bans on polygamy, same-sex marriage, and interracial marriage. The legal structure of intimate relationships (at least those formally recognized, such as marriage and family) is not itself private.

However, concern about the private sphere, construed as a zone of protected liberty around the person and in consensual relations, is abiding within feminism. In her landmark critique of privacy, Catharine MacKinnon wrote: “the very things feminism regards as central to the subjection of women … form the core of the privacy doctrine’s coverage …. The legal concept of privacy can and has shielded the place of battery, marital rape, and women’s exploited domestic labor.”74 The worry, then, is that privacy rights, or whatever liberties are construed as falling under the doctrine of privacy, prevent effective intervention by protecting choices which create vulnerability. This brings us back to the tradeoffs I noted above between protecting security and liberties (construed as part of the domain of privacy), particularly freedoms of speech, religion, and association. If socially created gender norms of dominance and entitlement produce IPV, then effective protection of security seems to require addressing gender norms (as the Resolve to Stop the Violence Program does).75 But effective interventions may be in tension with privacy.

These tensions are not easy to resolve, but I want to suggest how my account would handle them and how the priority of security might affect them. First, clearly security would not allow all kinds of interference with liberty, as I noted above. Freedom of association would protect the choice to associate, which entails that the state should not stop victims from returning to their abusers if they freely choose to (with real options to do otherwise).76 On the security approach, this seems right: there is no way to prevent association without restricting bodily movement, also imperiling security.

But the terms of intimate relationships are not completely shielded from interference, even if relationships are informal. Just as employer/employee relationships have defined protections built in law, so can intimate relationships (such as economic terms protecting exit). A couple may choose immersive, life-structuring forms of BDSM, living voluntarily in dominant and submissive roles, or as head of household and helpmeet, but their exit options should be protected and special legal protections against the vulnerabilities of access should be available to them. (Of course, a pressing concern well beyond the scope of this article is the transmission of such values to children.77)

Social pressures and expectations, as relayed and created through speech, media, religion, small associations, and so on, are another domain in which privacy, or liberties falling under privacy, seems to protect the creation of vulnerability. Many social and cultural products portray IPV in ways which normalize or even encourage it. Should parents be prevented from reading fairy tales which depict IPV to their children (we might be worried about Bluebeard, Beauty and the Beast, or Little Red Riding Hood, as well as much of Greek mythology)? Should violent pornography or video games be censored or otherwise regulated?78 Should the state intervene with ‘tradwife’ influencers who promote economic dependency for young women, potentially increasing their vulnerability to violence? And what of religious texts which depict domestic abuse as normal, and escape from it as against religious teachings? All of these could be construed as within the domain of privacy or protected liberties, yet all might seem to increase vulnerability to IPV.

The protected sphere of liberties will limit intrusions into speech and religion. But there is much the state can do consistent with respecting liberties. First, consider again the analogy of public health warnings about smoking. Intimate relationships create risk, and certain kinds create more risk. Public education can teach these risks, doctors can screen for risks, and the state can limit content (e.g. require warnings against children viewing content including images of smoking). The state can tax cigarettes and regulate their manufacturers, requiring warning labels. Recognizing the risk and disincentivizing risk-taking will affect behavior.

Second, protecting exit options (and providing protections against vulnerabilities of access) also affects decision-making. As noted above, women who have the means to leave are much more likely to leave abuse than women who do not, and economic dependency increases vulnerability to abuse. So it is at least prima facie plausible that even if social pressures lead women into abusive situations, if women have viable exit options, social pressures, including internalized gender norms, will not always keep them there. Obviously, the extent to which such interventions are effective is a vitally important empirical question. There is reason to believe that policies of the sort I am proposing would be effective in protecting security; whether they would be effective enough, and what should count as effective enough, are further questions.

In sum, not all state intervention in relationships is wrongful. The question is how the state should interfere, and why. On the account developed here, just as we have a claim to legal entitlements to protect us against the vulnerabilities which arise in employer-employee relationships, we also have a claim to legal entitlements to protect us against the vulnerabilities which arise in intimate relationships. As noted, similar vulnerabilities arise in employer-employee relationships: economic dependency, potential retaliation, surveillance, and control. The state should construct a secure context for intimate relationships, just as it should construct a secure context for employment. This need not be paternalistic, perfectionist, or privacy-violating.

The security-based argument for direct protections against violence has a number of benefits. Notably, because of the high priority of security, it justifies policies often defended on grounds of gender justice (because women are victims of IPV at higher rates) at a higher priority and with fewer controversial theoretical commitments than distributive arguments for similar policies (cf. Okin 1989, inter alia). These include policies such as subsidized childcare, financial and housing support on divorce or ending cohabitation, and access to health insurance independent of marriage.

As noted at the outset, interventions protecting against physical violence may also protect against emotional and other forms of abuse and control. Exit options and protections against vulnerabilities of access will protect victims of emotional abuse as well as physical abuse. Further, universal preventative interventions may improve the quality of existing relationships. Programs aimed at fostering lower-conflict relationships may improve the quality of existing relationships, thereby promoting goods of caring relationships which, arguably, the state should support.79

Indeed, this approach intervenes in recent debates over marriage law, particularly over marriage “abolition.”80 Clare Chambers, for example, has recently argued that the neutral state has no legitimate reason for recognizing intimate or caring relationships in marriage (or marriage-like) law.81 While I have previously argued that the value of such relationships can provide a reason for their recognition acceptable to the neutral liberal,82 this article provides a distinct line of argument for marriage recognition: marriage (or marriage-like law recognizing intimate relationships, such as civil unions or minimal marriage) serves to identify contexts of potential harm.

Intimate relationships are special sites of vulnerability to IPV. Thus, an institution recognizing and providing a legal framework for intimate relationships – like marriage – can flag these contexts of enhanced vulnerability. Legal recognition of contexts of vulnerability can be the basis for protective rights, as in employee-employer relationships. Likewise, marital status can flag these as sites for protective intervention. Security gives a neutral reason for recognizing marriage, or something like it, in law: because marriage is a site of violence, recognizing marriage flags sites of vulnerability for potential intervention. Basic security is a reason for recognizing relationships, because they are sites of vulnerability to violence.

One might worry that recognizing such relationships as a category will (illiberally) suggest their superiority and incentivize entry into them. Chambers writes that recognizing a status of relationships perpetuates a symbolic hierarchy: “the formalized, ceremonial act of naming some relationships and not others gives those named relationships a veneer of state-sanctioned respectability and approbation.”83 But recognizing relationships as sites of special vulnerability to violence does not endorse them – it is like adding red flags or warning labels to a dangerous activity. Presented thus, the status may urge caution, rather than incentivize forming highly vulnerable relationships.

Finally, the approach developed here coheres with non-carceral solutions to IPV proposed by anti-carceral feminists. Recent critics of “carceral feminism” have argued that criminal law enforcement disproportionately harms people of color, through police brutality, wrongful imprisonment, prison violence, and economic and political disenfranchisement.84 Anti-carceral feminists argue that mainstream feminists have wrongly turned to criminal law to protect women, in light of these harms and injustices.

Anti-carceral feminist arguments do not simply focus on the harm to alleged abusers in an often violent law enforcement system, but they also argue that criminalizing IPV and enforcing such law inadvertently harms victims of IPV, “disproportionately … women of color.”85 Such harms include victims being wrongly arrested, including for self-defense, wrongly incarcerated, being deprived of economic support by imprisoning their abusers, and suffering an increase in retaliatory violence.86 It is a benefit of my approach that its proposed interventions dovetail with concrete proposals made by anti-carceral feminists, such as shifting funding from law enforcement to support for victims.

Of course, my argument does not address the central issues in the debate over abolition. For prison abolitionists such as Angela Davis, the function of the prison system is to maintain white supremacy. Davis would argue that prison abolition is necessary to protect security: the carceral system creates vulnerability and exacerbates violence throughout society, including violence within prisons.87 As Erin Kelly writes in reply to Tommie Shelby’s defense of incarceration, for prison abolitionists such as Davis, arguments for reform do not grasp that its function is to maintain white supremacy, “mask[ing] and amplif[ying] the interaction of socioeconomic inequality, black poverty, and hostility toward black Americans by white Americans.”88 Prisons also produce IPV in prison in relationships of coerced sex for protection.89 So (the prison abolitionist reply would continue) we cannot effectively protect security without prison abolition. While I cannot address these arguments here, my argument for eliminating state-created vulnerability and protecting security throughout legal institutions extends to other security risks created by law and policy – such as brutality within the carceral system. Taking security seriously can help us rethink other areas of law and whether they adequately protect the security of all affected.

It is a benefit of my approach that my proposals dovetail with concrete proposals made by such non-ideal theorists, particularly abolitionist feminist proponents of transformative justice. Transformative justice, like restorative justice, aims at healing, reconciliation, and acknowledging the harm to the victim – but transformative justice also demands prison abolition.90 Shelby writes, “As abolition feminists emphasize, transformative justice is often a fitting response to domestic, gender, and sexual violence. In these cases, support for survivors, reconciliation, and healing from trauma are paramount, and penalizing those who caused the harm is secondary or sometimes counterproductive.”91

This article shows that elements of a restorative or transformative justice approach can be grounded within liberalism. I say “elements” of transformative justice, since transformative justice involves abolition. While I have not defended prison abolition, my approach to IPV incorporates restorative and transformative justice elements of support for victims and prevention of IPV. This is surprising, in part, because transformative justice emerges from radical views which are critical of liberal reformism.92 I have argued that the transformative justice goal of empowering victims – of protecting exit options and access to material conditions of security – is required by liberalism as a condition of the protection of basic liberties.

This article has asked how we would protect security if women’s security were taken seriously, making the kind of risks to which women are more vulnerable a focal point of security protections. Taking women’s security seriously has surprising implications. I have argued that intimate relationships create vulnerability to violence and as such are apt for intervention on grounds of security. In the context of IPV, prioritizing security has implications for an array of legal institutions, including marriage and family law, housing, healthcare, and education. The modest premise about basic security implies a set of reforms protecting victims’ exit options and protecting against vulnerabilities of access. This is what justice requires with regards to intimate partner violence.