Chief Justice DURRANT, opinion of the Court:

Introduction

¶ 1 This case requires us to consider the conditions under which the custodian of a dangerous person has a duty to prevent that person from injuring others. In prior cases, we have concluded that such a duty exists only if the custodian is aware, or should be aware, that the person poses a threat to a specific individual or a discrete group of individuals. In contrast, the Second Restatement of Torts does not require notice of the same particularized danger, and the Plaintiff in this case urges us to overrule our prior caselaw in favor of the Restatement's approach. For three reasons, we accept that invitation and adopt the standard articulated in the Restatement. First, our caselaw in this area is based on incorrect assumptions about the practical consequences of imposing such a duty. Second, Utah law is out of step with the rule employed in the overwhelming majority of other jurisdictions. And third, the old rule is inconsistent with the analytical framework we have employed in our most recent cases analyzing whether a defendant owes a duty of care.

¶ 2 We must also determine whether the Governmental Immunity Act as applied in this case violates article I, section 11 of the Utah Constitution (the open courts clause). We have read the open courts clause to prohibit the legislature from eliminating a cause of action unless it provides an alternative remedy that meets certain criteria. As we explain in more detail below, the Governmental Immunity Act grants governmental entities blanket immunity from any liability that arises from the exercise of a "governmental function." The legislature recently expanded the definition of that term to encompass any act or omission on the part of a governmental actor, and the Plaintiff in this case has asserted a tort claim against Utah County for its negligent operation of a prison work-release program. The parties concede that under the most recent version of the Governmental Immunity Act, the County is immune from suit. The question, then, is whether the legislature's expansion of governmental immunity eliminated a cause of action that the Plaintiff could have maintained against the County before the Act was amended. If it did, then the Act's application in this case may run afoul of the open courts clause.

¶ 3 We conclude that the Governmental Immunity Act is not unconstitutional as applied in this case. Even before the Act's expansion of immunity, its blanket immunity protections extended to any liability that arose from the performance of a uniquely governmental function or other acts that are essential to a core government activity. In this case, the Plaintiff's negligence claim arises directly from a prison work-release program. Because incarcerating and rehabilitating inmates falls squarely within that definition, the Act would have shielded the County from liability even if the Plaintiff brought suit before the legislature expanded blanket immunity protections to encompass a much wider range of activity. We affirm the district court's ruling on that basis.

Background

¶ 4 This appeal followed the district court's dismissal of Mika Scott's complaint against Utah County, Intermountain Employment Services (IES), and Universal Industrial Sales (Universal). On appeal from a district court's decision granting a motion to dismiss, we view the facts pled in the complaint and all reasonable inferences from them in the light most favorable to the plaintiff.1 We recite the facts consistent with that standard.

¶ 5 For some time, Utah County has operated a program known as "Jail Industries," which allows inmates to "work for private businesses in the community setting rather than on correctional institution grounds." The County actively seeks out private businesses to participate in the program, "emphasizing that by hiring inmates," the companies "assist in the rehabilitation of Utah County inmates, assist in the solvency of the Utah County budget, and receive a substantial discount on the price of labor." The County retains seventy-five percent of the inmates' earnings, and over the past decade, the program has "produced over $5,000,000 in gross revenues." Not all inmates are eligible for Jail Industries — the County screens each inmate that enlists in the program and does not place anyone it has not approved with a private employer.

¶ 6 IES worked with the County "to place" qualified inmates with private employers. In the past, "many" of these inmates "flagrantly disobeyed the rules they agreed to when enlisting" in the program, "walking away from the private jobsites" during the day, receiving illegal visits from friends and family, and using alcohol and drugs. But employers typically waited until the end of the work day to report these violations. Consequently, the County was aware that an inmate "could walk away from a private jobsite and the inmate's absence might not be noted for the better part of a day."

¶ 7 One of the inmates the County selected to participate in Jail Industries was Shawn Michael Leonard. IES placed Mr. Leonard with Defendant Universal in June 2010. According to Ms. Scott's complaint, the County "improperly screened" Mr. Leonard "for approval within the Jail Industries program because of a known potential for violent behavior toward other people" and "his extensive criminal history, which included a prior sentence in the Utah State Prison." Proper screening would have revealed that Mr. Leonard was "not eligible to participate" in the program and that he "posed" a particular danger "to young women living in the vicinity" of the work site.

¶ 8 But the County had "only one employee screen inmates and" did "not conduct[] one-on-one interviews with" Mr. Leonard or any other "inmates before placing them in the Jail Industries program." These improper screening procedures resulted in part from the County's efforts to increase revenue — that is, "the total number of inmates in the Jail Industries program was driven by the demand from the private businesses, not by the supply of qualified inmates." For their part, IES and Universal "knew or should have known that the participants in the Jail Industries program were actual inmates of Utah County, and that they were therefore not trustworthy and potentially dangerous to the public." The companies also "knew or should have known" that the inmates "regularly broke" program rules, "including walking away from the private jobsite and potentially committing crimes, and engaging in alcohol and ... drug use."

¶ 9 Mr. Leonard's participation in Jail Industries proved to be a tragic mistake. The County did not provide guards or any means of remotely supervising the inmates employed at Universal. And Universal failed to take any action to prevent the inmates from leaving the work site. As a result, on June 8, 2010, Mr. Leonard escaped. Universal did not report Mr. Leonard's absence until about one hour after his escape, and it took the County another hour to notify police that he had indeed left the work site.

¶ 10 The next day, Mr. Leonard approached Ms. Scott on the Provo River Trail about ten miles away from where he had been working. He grabbed Ms. Scott, covered her mouth, and told her not to scream. After forcing her off the trail into the bushes, Mr. Leonard strangled her with a shoe string. Ms. Scott soon lost consciousness, and Mr. Leonard then hit her repeatedly in the head with a cinder block, sexually assaulted her, and left. Ms. Scott survived, but her injuries were substantial. She had multiple surgeries to reconstruct her face and mouth; her jaw was wired shut for months; and she contracted a heart condition, permanent scars, anxiety, insomnia, and permanent double vision.

¶ 11 Ms. Scott filed a negligence action against the County, IES, and Universal in September 2011. She amended her complaint twice, and then all three Defendants moved to dismiss the second amended complaint. Ms. Scott opposed the dismissal and sought leave to file a third amended complaint. The district court ruled in favor of the Defendants, concluding that none of them owed a duty to Ms. Scott and denying her motion to amend as futile. As an alternative basis for dismissing the claims against the County, the district court also concluded that the Utah Governmental Immunity Act barred all of her claims against the County. Ms. Scott appealed.

¶ 12 Following oral argument in this case, Ms. Scott settled her claims against IES and Universal, but not the County. The parties to the settlement agreement then filed a suggestion of mootness under rule 37(a) of the Utah Rules of Appellate Procedure. We agree that Ms. Scott's claims against IES and Universal are now moot, and we accordingly dismiss them.2 We have jurisdiction over the remaining claims under Utah Code section 78A-3-102(3)(j).

Standard of Review

¶ 13 Ms. Scott argues that the district court improperly dismissed her negligence claim for failing to allege enough facts to establish a duty. We review a decision granting a motion to dismiss "for correctness, granting no deference to the decision of the district court."3 In so doing, we "accept the plaintiff's description of the facts alleged in the complaint to be true, but we need not accept extrinsic facts not pleaded nor need we accept legal conclusions in contradiction to the pleaded facts."4 Ms. Scott also argues that the application of the Utah Governmental Immunity Act "violates the open courts clause of the Utah Constitution." A constitutional challenge to a statute is a question of law, which we review for correctness.5

Analysis

¶ 14 In negligence cases involving "a defense of governmental immunity," we first determine "whether the defendant owed a duty of due care to the plaintiff before deciding whether the defendant is entitled to the affirmative defense of governmental immunity."6 We do so for both policy reasons and practical considerations. By analyzing questions of duty and immunity in that order, "a court can more clearly define the scope of each body of law and the policies that underlie them."7 And as a practical matter, if a governmental agency owes no duty of care, "there can be no prima facie case of negligence as a matter of law, and immunity would be immaterial."8

¶ 15 Accordingly, we first address whether the County owed Ms. Scott a duty of care and then discuss her open courts clause challenge to the Governmental Immunity Act. We conclude that the County did owe Ms. Scott a duty of care because it took affirmative steps that created a risk of harm — it established an off-site work-release program for potentially dangerous inmates in its custody and screened each inmate before placing them with employers.

¶ 16 But even though the County owed Ms. Scott a duty of care, governmental immunity bars her claim unless the application of the Governmental Immunity Act in this case violates the open courts clause of the Utah Constitution. We conclude, however, that the Act is not unconstitutional as applied in this case. We have read the open courts clause to prevent the legislature from eliminating a cause of action without providing an alternative remedy. In the context of governmental immunity, this means that any law expanding governmental immunity may violate the open courts clause by eliminating a claim a plaintiff could have brought against a governmental entity before the law's enactment. But here, the Act's application is not unconstitutional, because governmental immunity would have barred Ms. Scott's claim even before the legislature amended the Act to expand governmental immunity. Prior to the Act's expansion, its blanket immunity protections extended to any liability that arose from the performance of core governmental functions. And here, incarcerating and rehabilitating inmates is such a function. Consequently, the County is immune from suit, and the district court properly dismissed her complaint on that basis.

I. Duty and the Rollins Rule

¶ 17 Before analyzing whether the County owed Ms. Scott a duty, we first address Ms. Scott's argument that the controlling caselaw on the duty issue (the Rollins rule) should be overruled. After setting forth the rule and discussing its underlying policies, we overrule the Rollins rule because it is based on flawed reasoning, is out of step with the vast majority of other jurisdictions, and is inconsistent with our most recent negligence cases.

¶ 18 The district court determined that none of the Defendants owed Ms. Scott a duty and, accordingly, dismissed her negligence claim without discussing breach, causation, or damages. In so doing, the court applied the Rollins rule, which we articulated in three prior cases involving dangerous individuals who injured others during their release from a hospital or correctional facility.9 The Rollins rule provides that "[b]efore any duty is imposed to protect others from bodily harm caused by one" in the custody of another, "the `others' to whom such bodily harm is `likely' and in favor of whom the duty arises must be reasonably identifiable by the custodian either individually or as members of a distinct group."10

¶ 19 For example, in Ferree v. State, an inmate killed Mr. Ferree while on release from a community corrections center.11 The inmate had an extensive criminal history of non-violent property and drug crimes, and he was addicted to morphine, cocaine, and several other drugs.12 Prior to his release, the inmate received a psychological evaluation, which concluded that he "was an impulsive person who by his own admission acted without thinking and whose ready anger at even minor obstacles caused him to engage in antisocial acts."13 Despite these concerns, the corrections center approved the inmate's release to a halfway house and allowed him to leave for the weekend to attend a wedding.14 The inmate bludgeoned Mr. Ferree to death with a pipe two days later while intoxicated.15 Mr. Ferree's estate sued the corrections center for negligence.16 We affirmed a summary judgment ruling in favor of the corrections center, observing that there was no reason for officials "to suspect that [the inmate] was violent in general or would be violent toward a particular person or a particular type of person."17 And we concluded that for a duty to arise in these circumstances, officials must "have good reason to believe that a particular person may be jeopardized by the release of a prisoner who has demonstrated capacity for violence."18 We applied the same rule in Rollins v. Petersen19 and Higgins v. Salt Lake County,20 holding that no duty arises between the custodian of a dangerous individual and third parties unless the custodian is aware of a specific threat to the third parties that makes them "a potential target."21

¶ 20 Here, there are no allegations that Mr. Leonard planned to assault Ms. Scott prior to the attack. And according to our holding in Higgins, even Ms. Scott's allegations that he posed a particular danger to young women are not specific enough to create a duty.22 Thus, the Rollins rule required the district court to dismiss Ms. Scott's negligence claim — just like the victim in Ferree, Ms. Scott "was simply a member of the public, no more distinguishable to [the Defendants] than to any other person."23

¶ 21 Like any duty determination, the Rollins rule is a policy choice.24 In making that choice, we departed from the rule that appears to be followed in most jurisdictions,25 which imposes a duty of care on the custodian of a dangerous person if the custodian "knows or should know" that the person is "likely to cause bodily harm to others if not controlled."26 There is no requirement that the threat target a specific individual or distinct group of people.27

¶ 22 We justified our departure from the majority rule after weighing the importance of rehabilitative programs against the risk of injury to the public. Ultimately, we determined that the majority rule could threaten the future of such programs, "expos[ing] the state to potentially every wrong that flows from the necessary programs of rehabilitation and paroling of prisoners."28 Such a duty, we reasoned, would be "realistically incapable of performance," "closely approximate a strict liability standard of care,"29 and make "custodians running transitional programs virtual insurers of their services."30 Ms. Scott argues that the Rollins rule is inconsistent with our recent negligence caselaw and based on flawed premises, and she urges us to overrule it. We agree and overrule this line of cases.

¶ 23 "Those asking us to overturn prior precedent have a substantial burden of persuasion."31 To do so, we must be "clearly convinced that" prior caselaw "was originally erroneous or is no longer sound because of changing conditions."32 We also consider whether "substantial reliance interests ... counsel against overturning our precedent."33 We overrule the Rollins line of cases because, despite the reliance interests of hospitals and correctional facilities, (1) the Rollins rule was based on dubious assumptions when it was decided, (2) a strong majority of other states follow the Restatement rule, and (3) the duty analysis in Rollins is inconsistent with our recent caselaw.

¶ 24 First, the policy reasons we cited as support for the Rollins rule cannot withstand careful scrutiny. As we have discussed, Rollins's underlying premise is that prison officials and health care providers are incapable of preventing dangerous individuals in rehabilitative programs from harming members of the public, so imposing a duty to control them exposes the operators of such programs to massive liability.34 And because the principal beneficiaries of rehabilitative programs are often dangerous individuals, the mere recognition of a duty would impose a standard of care "realistically incapable of performance" and "fundamentally at odds with the nature of the parties' relationship."35 In other words, according to Rollins, recognizing a duty of care in this context would amount to strict liability whenever someone in a rehabilitative program injures a third party, effectively bankrupting these important programs.36

¶ 25 That conclusion is not consistent, however, with basic principles of tort law, which limit liability even when a duty exists. Negligence claims have four distinct elements — duty, breach, causation, and damages.37 The question of whether a duty exists is therefore analytically distinct from whether the defendant in a particular case acted reasonably enough to meet the applicable standard of care.38 And as every first-year law student learns, a defendant who takes reasonable precautions to prevent injury can avoid liability, if, notwithstanding her best efforts, she nevertheless injures the plaintiff.39 It is therefore simply not correct that recognizing a duty would, in practice, impose liability on the custodian of a dangerous individual every time the individual harmed a third party.

¶ 26 For instance, if a prison carefully screened participants in work-release programs to assure that only model inmates with no history of violence could participate, it would be difficult to say the prison breached a duty if an inmate escaped while on release and committed a violent crime. And even when inmates with violent criminal histories participate in work-release programs, rigorous screening procedures and appropriate supervision may be reasonable steps that could prevent many injuries without imposing prohibitive costs.

¶ 27 Moreover, other negligence principles further limit a custodian's potential liability. Under the proximate cause element, prison officials and hospitals cannot be liable unless the plaintiff's injuries are a foreseeable result of their negligence.40 Thus, the greater the temporal and geographic distance between the plaintiff's injury and the custodian's role in releasing the dangerous individual, the more difficult it is for plaintiffs to establish a causal link between their injury and any breach of duty.41

¶ 28 Not only is the reasoning in our prior caselaw questionable, but a strong majority of states impose a duty much broader than the Rollins rule on custodians of dangerous individuals. Thirty jurisdictions require hospitals and prisons to protect third parties from dangerous people in their custody even if the custodian is not aware of a threat to a specific individual or group. These jurisdictions have either explicitly adopted section 319 of the Second Restatement of Torts42 or cited the Restatement approvingly and followed its general approach.43 Section 319 provides that "[o]ne who takes charge of a third person whom he knows or should know to be likely to cause bodily harm to others if not controlled is under a duty to exercise reasonable care to control the third person to prevent him from doing such harm."44 While there are subtle differences between the Restatement and how states have chosen to implement it, very few require the person with custody of the dangerous individual to be aware of threats to a specific victim or group that includes the victim. Aside from Utah, we have found just four states that impose such a requirement.45 And as we discuss later, almost all of the jurisdictions that follow the Restatement's approach also operate rehabilitative programs.46

¶ 29 Finally, our most recent negligence caselaw is more consistent with the Restatement than it is with Rollins. In B.R. ex rel. Jeffs v. West, we identified a number of factors that are "relevant to determining whether a defendant owes a duty to a plaintiff."47 These include "(1) whether the defendant's allegedly tortious conduct consists of an affirmative act or merely an omission; (2) the legal relationship of the parties; (3) the foreseeability or likelihood of injury; (4) public policy as to which party can best bear the loss occasioned by the injury; and (5) other general policy considerations."48 We made clear that each factor must be "analyzed at a broad, categorical level for a class of defendants"49 rather than a factually intense inquiry "decided on a case-by-case basis."50 If "the relevant category of cases" "includes individual cases in which the likelihood of some type of harm is sufficiently high that a reasonable person could anticipate a general risk of injury to others," the defendant likely owes a duty of care to the plaintiff.51

¶ 30 In the cases in which we adopted and applied the Rollins rule, we focused on specific factual considerations to determine whether a duty existed rather than examining the parties' relationship in broad categorical terms. In Ferree, we determined that a corrections center owed no duty of care because the plaintiffs "presented no evidence that Ferguson" — a temporarily released inmate — "had previously exhibited violent behavior toward another or that he had physically threatened another."52 We observed that even though "proof of the allegations" in the plaintiffs' complaint "might establish lack of due care in the abstract," there was "nothing to indicate that the officials were aware of anything more than a generalized possibility" of harm.53 Similarly, in Rollins we concluded that a hospital did not owe a duty to a motorist injured by an escaped mental patient because the "record [was] devoid of any evidence" that the patient had "set himself apart in terms of dangerousness to [the motorist] personally or to any distinct group of which [the motorist] was a member."54

¶ 31 This type of specific, case-by-case analysis is incompatible with our directive in Jeffs that courts articulate a party's duty of care "in relatively clear, categorical, bright-line rules of law applicable to a general class of cases."55 In Jeffs, we clarified that the "essential difference among the elements [of negligence] is that duty is a question of law determined on a categorical basis, while breach and proximate cause are questions for the fact finder determined on a case-specific basis."56 In adopting and applying the Rollins rule, we therefore conflated the case-specific analysis properly reserved for breach and causation with the abstract, categorical inquiry that should have been employed to articulate the duty.

¶ 32 In sum, the Rollins rule is based on flawed premises, inconsistent with the law in most other jurisdictions, and at odds with our own most recent negligence caselaw. Consequently, even though departing from the rule may upset the reliance interests of correctional facilities and health care providers that regularly house dangerous individuals, we overrule Rollins.

II. Under the Proper Duty Analysis, We Conclude that the County Owed Ms. Scott a Duty and Adopt Section 319 of the Second Restatement of Torts

¶ 33 Having overruled the Rollins rule, we now consider anew the circumstances under which the custodian of a dangerous person owes a duty to third parties that the dangerous person injures. As we have just discussed, Jeffs requires tort duties to be articulated "in relatively clear, categorical, bright-line rules of law applicable to a general class of cases."57 We therefore analyze each pertinent factor in the duty analysis "at a broad, categorical level for a class of defendants" without focusing on the particular circumstances of a given case.58 These factors are "(1) whether the defendant's allegedly tortious conduct consists of an affirmative act or merely an omission; (2) the legal relationship of the parties; (3) the foreseeability or likelihood of injury; (4) public policy as to which party can best bear the loss occasioned by the injury; and (5) other general policy considerations."59 Below, we analyze each of these factors in turn. In so doing, we adopt section 319 of the Second Restatement of Torts and hold that the custodian of a dangerous individual has a duty of care to prevent that individual from harming members of the public. And under this standard, we conclude that Ms. Scott pled enough facts to establish that the County owed her a duty.

¶ 34 Each of the five duty factors we articulated in Jeffs favors imposing a duty on the County. First, operating a work-release program is an affirmative act, not an omission. Second, while the County had no legal relationship with Ms. Scott, it did have a custodial relationship with her attacker. Third, failing to adequately screen inmates before allowing them to participate in a temporary work-release program could foreseeably result in dangerous individuals harming others. Fourth, it is the custodian of the dangerous individual — not potential victims — that is best situated to bear the loss associated with such an injury. And finally, numerous jurisdictions impose a duty on prisons and hospitals to control dangerous individuals in their custody, and all of them operate transitional rehabilitative programs.

¶ 35 The first duty factor favors imposing a duty of care on the County because screening and placing inmates in a work-release program is an affirmative act, not an omission. The distinction between passive inaction and affirmative acts is "central to assessing" the duty question,60 because the law imposes a duty of care "where an individual's active misconduct work[ed] positive injury to others."61 But an omission, or the "failure to take positive steps to benefit others," gives rise to a duty only when there is a special legal relationship between the parties.62 The line between acts and omissions is sometimes subtle. Borrowing from Justice Cardozo, we have characterized the inquiry as "whether the putative wrongdoer has advanced to such a point as to have launched a force or instrument of harm, or has stopped where inaction is at most a refusal to become an instrument for good."63

¶ 36 The conduct in this case involves more than the passive failure to protect another. When the custodian of potentially dangerous individuals negligently places those individuals in a rehabilitative program, that action "launche[s] a force or instrument of harm,"64 creating a risk of injury to others. And here, the County created a work-release program and placed potentially dangerous inmates with private companies. It screened each inmate who volunteered, refusing to place anyone with an employer who was not approved. And once approved, the inmates worked outside prison walls without any meaningful supervision by prison officials. By placing inmates in the community, the County engaged in "active misconduct" if its screening procedures were inadequate to discover obvious dangers work-release participants might pose to the public. Consequently, this is not a case where liability stems from a plaintiff's failure to warn or take other affirmative steps to protect the defendant. Instead, the County's conduct involves an affirmative act, and this factor accordingly weighs in favor of imposing a duty.

¶ 37 Next, we turn to the second factor in our duty analysis — the parties' relationship. This factor favors imposing a duty if there is a special legal relationship between the parties.65 Some examples include "common carrier to its passenger, innkeeper and guest, landowner and invitee to his land, and one who takes custody of another."66 The County argues that Ms. Scott must show a special relationship to establish a duty, most likely because historically the public duty doctrine has prohibited imposing a duty on governmental actors for even affirmative acts absent a special relationship.67 But in Cope v. Utah Valley State College, a case issued last year, we made clear that the public duty doctrine no longer imposes such a requirement when a governmental defendant's negligence stems from an affirmative act rather than an omission.68 Here, because the County's negligence arises from an affirmative act, Ms. Scott does not need to establish a special relationship between herself and the County for her claim to survive dismissal.

¶ 38 But even if she did, there is a legal relationship that favors imposing a duty. We have recognized that someone who has "actual custody ... of a third person who causes harm to the plaintiff"69 may have a duty if he knew, or should have known, that the third person is dangerous. Two aspects of the custodial relationship between prison officials and inmates show why such a duty should apply in this case. First, prison70 officials "exert actual, physical dominion and control over the prisoners,"71 giving them a wealth of information about the inmates' physical and mental capacities to which no one else has comparable access. Second, such a relationship involves the "legal authority to control" the person's movement and interactions with the public.72

¶ 39 The custodian's knowledge should inform the manner in which it exercises control — a reasonable custodian that knows an individual is dangerous would impose more constraints on that individual than someone without violent tendencies. And the custodian would also take reasonable steps to discover this information before making decisions that could expose the individual to potential victims. Section 319 of the Second Restatement of Torts articulates just such a duty: "One who takes charge of a third person whom he knows or should know to be likely to cause bodily harm to others if not controlled is under a duty to exercise reasonable care to control the third person to prevent him from doing such harm."73 By way of illustration, the Restatement provides an example that is particularly apt to the facts of this case: "A operates a private sanitarium for the insane. Through the negligence of the guards employed by A, B, a homicidal maniac, is permitted to escape. B attacks and causes harm to C. A is subject to liability to C."74

¶ 40 The County argues that its relationship with Mr. Leonard is insufficient to create a duty. It maintains that a custodian's duty should hinge on whether it "has actual, physical control over the individual." And here, Mr. Leonard was not under the County's direct supervision when he escaped.

¶ 41 Accepting the County's argument, however, would create perverse incentives. If a custodian's duty in this context were limited solely to protecting others from dangerous people under its actual, physical control at the time of an attack, prisons and hospitals that remain willfully blind to an individual's violent tendencies when releasing him or her into a rehabilitative program would be placed in precisely the same position as a custodian who took every precaution to ensure no dangerous individual was temporarily released — neither would face liability. By contrast, extending a custodian's duty to the manner in which it exercises legal control imposes greater liability on custodians who fail to screen inmates or patients for violent tendencies than it does those who employ prudent screening procedures. Focusing the duty on legal control therefore properly encourages all custodians to be careful about which individuals they expose to the general public through rehabilitative programs.

¶ 42 For these reasons, we adopt the Restatement standard for determining when the custodian of a dangerous individual owes a duty to prevent the individual from injuring others. And we conclude that because the County had legal custody of Ms. Scott's attacker, this factor weighs in favor of imposing a duty. We now turn to the next factor, foreseeability.

¶ 43 This factor, the third in our duty analysis, also favors imposing a duty on the County. As discussed earlier, foreseeability analysis for duty purposes differs in kind from foreseeability in proximate cause.75 We do not examine whether the County could have foreseen the specific chain of events that led to Ms. Scott's injury. Rather, we ask "whether a category of cases includes individual cases in which the likelihood of some type of harm is sufficiently high that a reasonable person could anticipate a general risk of injury to others."76 Here, the relevant category of cases involves the custodian of a potentially dangerous individual who places the individual in the community out-side its direct physical control with minimal supervision. And in so doing, the custodian fails to adequately evaluate the individual for potential dangerousness.

¶ 44 There are certainly circumstances within this class of cases in which the custodian could foresee a risk of injury. For example, inmates are in state custody. Some are nonviolent offenders who would pose little, if any, foreseeable danger to the public if temporarily released. Others may have committed violent offenses but have been model citizens throughout their prison term. But there are also other categories of inmates who have significant disciplinary problems in prison, a history of violence, mental illnesses, problems with substance abuse, or a combination of several of these issues. Inmates in this category pose a heightened risk of harm to others if allowed to work outside the prison without meaningful supervision. Consequently, a custodian that employs inadequate screening procedures could certainly foresee one of these dangerous individuals escaping from a minimally supervised work site and harming someone. The foreseeability factor therefore favors imposing a duty on the County.

¶ 45 The fourth duty factor — "public policy as to which party can best bear the loss occasioned by the injury" — similarly supports imposing a duty of care on the County.77 Under this factor, we do not consider "the depth of [the parties'] pockets," but instead examine which of them "is best situated to take reasonable precautions to avoid injury."78 In Jeffs, we determined that physicians are in the best position to prevent injuries to third parties caused by patients who received erroneously prescribed medication.79 We noted that medical experts "can take into account the propensities of the drug, as well as the susceptibilities of [the] patient," and we concluded that "the prescribing physician of a prescription drug is the person best able to take or recommend precautions against potential injuries."80

¶ 46 Our reasoning in Jeffs applies with equal force here. Although it is true that private employers and potential victims can access public records detailing an inmate's criminal history, only prison officials are acquainted with the inmate's behavior since his conviction. Prison officials' daily interactions with inmates give them important insights about who can be trusted to participate in a temporary work-release program and who should remain behind bars. Just as a physician becomes acquainted with a patient's particular response to different medications, prison officials become intimately familiar with which inmates routinely abuse privileges and create conflict. No one can predict with perfect accuracy whether an inmate will injure someone during temporary release,81 but it is difficult to imagine anyone in a better position to assess that risk than the custodian charged with supervising the inmate on a daily basis. Accordingly, this factor weighs in favor of imposing a duty.

¶ 47 We now turn to the final factor — "other general policy considerations" — and conclude that the competing public policies at issue supports the imposition of a duty.82 We reach this conclusion because we see no reason why a policy favoring rehabilitative programs cannot coexist with the tort law policy of compensating injured parties. Both policies are important. With respect to rehabilitative programs, we have stated that "parole and minimum security programs are designed to give the inmate the best opportunity to successfully become a member of society again."83 So even though there is substantial "imprecision associated with predicting violent human conduct," such programs are "practically indispensable."84 We agree with those statements and reaffirm them today.

¶ 48 But recognizing a duty promotes equally weighty policy concerns, and we are not convinced that doing so would impede the State's ability to maintain rehabilitative programs. To begin with, the basic purpose of tort law is "to place an injured person in a position as nearly as possible to the position he would have occupied but for the defendant's" tortious behavior.85 Providing such compensation both "protect[s] societal interests in human life, health[,] and safety"86 and deters harmful behavior by requiring individuals whose conduct harms those around them to bear the full cost of their actions.87 Here, imposing a duty on custodians of dangerous individuals serves these important interests. A plaintiff, like Ms. Scott, can be compensated for her injuries, and the potential liability for prison officials and other custodians provides a powerful incentive to screen program participants rigorously.

¶ 49 Of course, subjecting prisons and hospitals to liability also raises the cost of rehabilitative programs, which appears to have been the Rollins court's central concern.88 But it is far from clear to us that these additional costs will be prohibitive — according to the Utah Sheriffs' Association's amicus brief, "every state except New Jersey [has] a formal temporary release program in place." And yet, at least thirty of these jurisdictions impose a duty similar to the Restatement standard.89 Moreover, the Sheriffs' Association also reports that in a survey of twenty-five counties in Utah, around 530 inmates are released to work in the community and 212 are released for employment or education. And of the twenty-five counties surveyed, just two "reported a negative interaction with the community." If the vast majority of these programs successfully place inmates without incident, we struggle to see why recognizing a duty endangers their fiscal integrity, particularly in light of other limiting principles of tort law we have already discussed.90 Thus, rather than undermining the State's interest in rehabilitating inmates, recognition of a duty in this instance accommodates the competing policies at stake.

¶ 50 In sum, each of the five factors we analyze to establish a duty of care favors imposing one on the County. We therefore conclude that the County owes Ms. Scott a duty of care and adopt the standard set forth in section 319 of the Second Restatement of Torts. The custodian of a dangerous individual must exercise reasonable care when deciding whether to allow that individual to participate in temporary release programs. And if the custodian's negligence allows a dangerous individual to harm someone while on release, the custodian may be liable for the harm.91

III. Even Though the County Owes Ms. Scott a Duty, It is Immune From Suit

¶ 51 Having concluded that the County owes Ms. Scott a duty, we now discuss whether it is immune from suit. We begin by noting that Ms. Scott concedes that the Governmental Immunity Act bars her claim. But she argues that the Act is unconstitutional as applied under article I, section 11 of the Utah Constitution (the open courts clause). That clause provides,

¶ 52 We have interpreted the open courts clause to prevent the legislature from passing a law that "abrogates a cause of action existing at the time of [the law's] enactment" unless it (1) provides "an effective and reasonable alternative remedy" or (2) "seeks to eliminate a clear social or economic evil" by means that are not "arbitrary or unreasonable."92

¶ 53 None of the parties has argued that Ms. Scott was afforded an alternative remedy, so the central question is whether the legislature abrogated her cause of action when it expanded the Governmental Immunity Act. Because the interaction of governmental immunity and the open courts clause is somewhat complex, we first briefly discuss the historical development of governmental immunity in Utah and then set forth the legal standard we apply for open courts clause challenges in this context. Applying that standard, we then conclude that the legislature did not abrogate a cause of action Ms. Scott would have had before it expanded governmental immunity, so the Act's application in this case is not unconstitutional.

A. Governmental Immunity and the Open Courts Clause

¶ 54 To determine whether the Governmental Immunity Act violates the open courts clause in a particular case, we look to see whether the plaintiff could have brought his or her cause of action prior to 1987.93 As we explain below, this is because 1987 is the high-water mark of governmental liability in Utah, and any subsequent amendments to the Act expanding governmental immunity therefore eliminates causes of action that could have been maintained against governmental entities before 1987.

¶ 55 Prior to the enactment of the Governmental Immunity Act in 1965, the common law doctrine of sovereign immunity prevented a citizen from suing a state governmental entity for any act considered to be a function of government.94 Our cases characterized a state action as "governmental" if it was not serving a "proprietary" function. By "proprietary," we meant that, in performing the action, the State obtained a pecuniary benefit, competed directly with private entities in the marketplace, or engaged in activity that could be successfully operated by private enterprise.95

¶ 56 The 1965 Governmental Immunity Act expanded liability for state entities beyond common law sovereign immunity by making the government subject to suit when it engaged in specific activities.96 It also provided that governmental entities retained blanket sovereign immunity protections when "engaged in the exercise and discharge of a governmental function."97 But nowhere in the Act did the legislature define the term "governmental function," so Utah courts relied on the governmental-proprietary function test from our sovereign immunity caselaw to interpret the full scope of immunity under the Act.98

¶ 57 In Standiford v. Salt Lake City Corporation, we expressly disavowed this precedent because it led to "contrary and unpredictable results."99 We held "that the test for determining governmental immunity" under the Act "is whether the activity under consideration is of such a unique nature that it can only be performed by a governmental agency or that it is essential to the core of governmental activity."100

¶ 58 Perhaps in response to our decision in Standiford, the legislature restricted governmental liability in 1987 by expanding the Act's definition of "governmental function" to include "any act, failure to act, operation, function, or undertaking" regardless of whether the activity "is characterized as governmental, proprietary, a core governmental function, unique to government, undertaken in a dual capacity, essential to or not essential to a government or governmental function, or could be performed by private enterprise or private persons."101 The legislature has continued to expand the definition of "governmental function" in subsequent amendments, and the statute currently defines that term as encompassing anything the government decides to do — "each activity, undertaking, or operation performed by a department, agency, employee, agent, or officer of a government entity."102

¶ 59 Anytime the legislature expands the definition of "governmental function," it restricts the government's liability beyond the scope of the Act as interpreted in Standiford — possibly abrogating causes of action that would have existed before the 1987 amendment and violating the open courts clause. Consequently, we have looked to the Standiford test — which defined "governmental function" in the Act before the legislature expanded the definition of that term in 1987 — to determine "whether the Act in its current form abrogates a cause of action that existed prior to its enactment."103

B. The Application of the Governmental Immunity Act in this Case is Not Unconstitutional

¶ 60 We now apply the Standiford test to resolve Ms. Scott's open courts clause challenge to the Governmental Immunity Act. As we have discussed, she must show that the legislature abrogated a cause of action she could have brought before 1987.104 To make that determination, we apply the Standiford test, assessing "whether the activity giving rise to the cause of action [1] is of such a unique nature that it can only be performed by a governmental agency or ... [2] is essential to the core of governmental activity."105 If operating Jail Industries falls under either category, then the County would have been immune from Ms. Scott's suit in 1987 before the legislature expanded governmental immunity, and the Act's application in this case would not violate the open courts clause.

¶ 61 The first category "does not refer to what government may do, but to what government alone must do."106 The second encompasses "those activities not unique in themselves (and thus not qualifying under the first part) but essential to the performance of those activities that are uniquely governmental."107 We have previously determined that the following activities were not governmental functions under Standiford: operating a public golf course,108 operating a public sledding hill,109 collecting and disposing of sewage,110 operating a municipal electrical power system,111 and operating a hospital where only "3.5 percent of the hospital's operating budget came from legislative appropriations."112 By contrast, we have held that transporting students to an out-of-state debate tournament,113 the regulation of boxing matches,114 and operating a public transportation system were governmental functions.115 In these cases, we identified the following factors as characteristics that weigh in favor of finding that an activity is a governmental function — the extent to which the activity is funded by the State, competes in the marketplace with private entities, generates annual profits, and would be "qualitatively different" if engaged in by a private entity.

¶ 62 Under this standard, we conclude that rehabilitation programs like Jail Industries are essential to the core governmental activity of running a state prison system. We have described rehabilitative programs for inmates as "necessary programs" that are "practically indispensable"116 to managing the prison population. Housing and rehabilitating inmates is an integral piece of the justice system, and if administering justice to those who violate the penal code is not a governmental function, we do not know what is.

¶ 63 Ms. Scott nevertheless argues that Jail Industries is qualitatively different than traditional work-release programs because "inmates, rather than parolees, were inserted into the community with little supervision as a source of revenue for Utah County and its private partners." Although it is true that a governmental activity that generates profits is more likely to be classified as a nongovernmental function under Standiford, profitability alone is not a determinative factor.117

¶ 64 Moreover, the fact that Jail Industries places inmates with employers outside the prison is insufficient to transform what we have recognized as a core governmental activity into a private endeavor. An activity that supports a core governmental function may satisfy the Standiford test even if it is not indispensable.118 For example, in Tindley v. Salt Lake City School District, we concluded that an extracurricular school debate program was essential to the core governmental function of educating students.119 We so held even though the negligence claim at issue arose from the school's efforts to transport the debate team to out-of-state competitions.120 We observed that the debate program "clearly benefits student education and is unlikely to be available to public school students if not offered through their schools."121

¶ 65 Similarly, Jail Industries gives inmates the benefit of work experience and a modest paycheck — significant experience that may ease their transition back into society and serves the core governmental function of rehabilitating inmates. It is certainly possible to house inmates without a program like Jail Industries. But the program "clearly benefits" inmates, and its unique benefits are "unlikely to be available" to them if the prison does not provide it.122

¶ 66 We therefore conclude that Jail Industries is essential to the core governmental function of housing and rehabilitating inmates, and the program accordingly qualifies as a "governmental function" under Standiford. Consequently, the County has always enjoyed immunity for such an activity, and the legislature's expansion of governmental immunity in 1987 did not abrogate Ms. Scott's cause of action. Accordingly, the Governmental Immunity Act is not unconstitutional as applied in this case.

Conclusion

¶ 67 We overrule the Rollins rule and hold that the custodian of a dangerous individual has a duty to take reasonable precautions to prevent that individual from injuring others. Under that standard, we conclude that the County owed Ms. Scott a duty. But even though the County owed Ms. Scott a duty, her negligence claim is barred by the Governmental Immunity Act. Finally, because work-release programs are essential to the core governmental activity of housing and rehabilitating inmates, the Act is not unconstitutional as applied in this case. We therefore affirm the district court's decision dismissing Ms. Scott's negligence claims against the County.