Similarly, a contributory-negligence defense may require reexamination of military decisions if it requires considering the fault of a military decision-maker.59 After a power outage, a Marine was electrocuted while trying to install a backup *257generator.60 Despite being told not to turn on the main generator because a group of Marines was working on it, the contractor turned it on, resulting in the Marine's injuries.61 The contractor asserted a contributory-negligence defense, which would have required the court to decide whether the Marines were reasonable in trying to install the additional generator and whether backup power should have been supplied to that area.62 The defense made the claim nonjusticiable.63
C
Even if a claim requires reexamination of a military decision, that decision must be one that is "insulated from judicial review."64 "[D]ecisions as to the composition, training, equipping, and control of a military force are essentially professional military judgments" beyond judicial review.65 Some decisions, such as whether to employ military force or the proper tactics to use during combat, are clearly professional military judgments that are beyond the judiciary's competence.66 But other decisions, at first glance appearing to be decisions similar to those civilians make and subject to judicial review, are still yet "complex[,] subtle, and professional decisions"67 that bear on military strategy.
As one circuit court has observed, "[h]ousing and maintenance decisions on a battlefield are exactly this type of decision".68 In the case of the soldier electrocuted while showering, the contractor's proportionate-liability defense injected a political question into the case.69 The fact-finder would be required to review "the military's decisions to house troops in unsafe barracks that would not be repaired."70 To choose which barracks in which to house troops and whether to repair them, the military must consider issues unique to the battlefield such as the danger relative to other options and the cost of repair relative to other uses of its scarce resources. Judges lack not only the constitutional authority but also the expertise to evaluate these decisions.
III
We turn now to the "discriminating analysis"71 required to apply these considerations in the political question doctrine to this case.
AMK9 contends that the Department of Defense (with which it contracted to provide Kallie) and the Army (in which Kallie served) caused Freeman's injuries, in part because the Army built (and rebuilt) Kallie's kennel to comply with its regulations and required AMK9 to use it. The trial court granted AMK9's motion to designate *258the Department and the Army as responsible third parties, thereby requiring "[t]he trier of fact ... [to] determine [their] percentage of responsibility ... for ... causing or contributing to cause" Freeman's injury.72 Thus, as one court has observed, the political question doctrine "loom[s] large."73
Freeman argues that the Department and the Army cannot be joined as responsible third parties because, as alleged by AMK9, their only duty to construct dog kennels was contractual and does not supply a sufficient legal standard for determining an allocation of responsibility to them. We think the contractual duty is sufficient, but in any event, the Army undertook to build Kallie's kennel and remediate it and required Kallie to use it. Had the actor been a private entity rather than the Army, these facts alone would support a negligent-undertaking claim.74 Therefore, we must decide whether litigating this case, including AMK9's proportionate-responsibility defense, will necessarily require reexamination of sensitive military decisions.
The military had plenary control over at least some of the decisions implicated by Freeman's claim. AMK9's contention that Freeman's alleged injury occurred when Kallie jumped over an internal partition between pens and escaped through an adjacent pen's open door calls into question the Army's design decisions not to extend the internal partition to the ceiling and not to cover the kennel. The Army designed and constructed the kennel and required AMK9 to use it. While Freeman argues that only AMK9's negligent failure to train and control Kallie caused her injury, AMK9 argues, and will argue at trial, that the Army's design was to blame. AMK9's proportionate-liability defense requires the fact-finder to evaluate these decisions. The Army's design decisions would be front and center at trial. On that point, this case is virtually indistinguishable from Harris , in which the United States Court of Appeals for the Third Circuit held that a private contractor's proportionate-liability defense would render the case nonjusticiable.75 If this case were to proceed, the fact-finder would be required to determine the degree to which the Army was responsible for Freeman's injury. This inquiry would require a reexamination of Army decisions, contrary to Baker 's first factor.
The Army's decisions about designing and constructing the kennels are unreviewable military decisions because they go to the equipping of the military, constitutionally committed to the federal political branches. This includes decisions about base configuration, including the design for kennels that house trained explosive-detection dogs like Kallie. Such decisions are similar to decisions about the quartering of soldiers and require similar risk-weighing judgments and allocation of scarce resources. Here, it is undisputed that the Army did not comply with its internal requirement to construct the kennel in a certain way, and a court should not insert itself into determining whether the Army should or should not have followed its guidelines. Were the roofs left off in order to allow the dogs to escape in the event of an attack? Or for ventilation in the desert heat? Or because those responsible for *259construction were summoned to other tasks? Or to conserve budgetary or material resources for additional structures? Only the Army can answer these questions in accounting for the construction of Kallie's kennel.
JUSTICE DEVINE argues that whether the Army actually caused Freeman's injury is a disputed issue of fact, which can be decided only by the jury.76 Of course, we agree. But the dissenting JUSTICES argue that the determination of whether to apply the political question doctrine must await full discovery and the jury's verdict.77 If the Army is found not to have caused Freeman's injury, then the case went to trial as it should have, and if the Army is found to have caused Freeman's injury, then the case should have been dismissed for want of jurisdiction-which no longer matters. Whether the political question doctrine applies or not, in the dissenting JUSTICES ' views, the case must be tried. The doctrine is reduced to an irrelevance. But the doctrine does not protect against determining the Army's liability. No one argues that the Army can be liable for Freeman's injury. Rather, the doctrine protects against judicial reexamination of military decisions.78 At least AMK9's defenses, and perhaps even Freeman's claim, cannot be adjudicated without putting the Army's conduct and decisions on trial. The political question doctrine requires us to be mindful of the broader implications of reviewing sensitive military decisions, such as maintaining respect for the separation of powers and the federalism system outlined in the United States Constitution, minimizing interference with military prerogatives, limiting military expenditures on participating in lawsuits (such as in discovery requests), and reducing contractors' liability for the sake of future procurement efforts.79 The inextricable involvement of military decisions in this case is not a matter of fact but a matter of law.
The dissenting JUSTICES argue incorrectly that our analysis ignores our usual process for deciding jurisdictional issues.80 As we have explained,
[w]hen the consideration of a trial court's subject matter jurisdiction requires the examination of evidence, the trial court exercises its discretion in deciding whether the jurisdictional determination should be made at a preliminary hearing or await a fuller development of the case, mindful that this determination must be made as soon as practicable.81
In this case, the trial court properly exercised its discretion to dismiss the case early on. "If the evidence creates a fact question regarding the jurisdictional issue , then the trial court cannot grant the plea to the jurisdiction, and the fact issue *260will be resolved by the fact finder."82 The dissenting JUSTICES assume that the issue in determining whether to apply the political question doctrine in this case is whether, in fact, the Army is responsible for causing Freeman's alleged injury. But the issue is not whether the Army can be held liable; all agree that it cannot be. The jurisdictional issue is whether litigating the case inextricably involves reviewing military decisions. It certainly does. The dissenting JUSTICES ignore "the fundamental precept that a court must not proceed on the merits of a case until legitimate challenges to its jurisdiction have been decided."83
JUSTICE DEVINE 's dissent also argues that today's decision "bars all tort suits where a military contractor-or any other defendant-is able to muster a mere allegation that a government actor whose decisions are insulated by the political-question doctrine partly caused the alleged harm."84 This is simply not true. The cases we have cited differentiate among claims in which military decisions are or are not inextricably involved. If, for example, Kallie bit Freeman while being routinely exercised by her civilian-contractor handler, her biting Freeman would have had nothing to do with the military. The political question doctrine is not always easy to apply, but it certainly cannot be invoked to bar all claims that merely happen to have a military setting.
Whether the Army was justified in ignoring its requirements and constructing the kennel as it did is not a question a Texas court can answer. Thus, we hold that this case is nonjusticiable due to the presence of an inextricable political question. We need not consider the other grounds that AMK9 asserts for dismissal.
IV
Hill Country did not join AMK9's plea to the jurisdiction or file its own. The court of appeals held that the trial court's sua sponte dismissal of the claims against Hill Country was erroneous because Hill Country did not submit authority to establish either Hill Country's immunity or the trial court's lack of subject matter jurisdiction. "Subject matter jurisdiction is an issue that may be raised for the first time on appeal[,] it may not be waived by the parties",85 and it may-indeed, must -be raised by an appellate court on its own.86 The political question doctrine examines justiciability, a jurisdictional matter. Thus, it may be raised at any time or by the court sua sponte .
Freeman's claims against Hill Country must be dismissed on the same political question grounds outlined above. Pragmatically, the case will almost certainly require examination of the same apportionment-of-liability questions outlined above, though up to this point Hill Country has not had to join or separately file a motion to designate a responsible third party. Because we favor early resolution of justiciability issues, we hold that the trial court was correct to dismiss Freeman's claims against Hill Country.
* * * * *
The district court correctly concluded that Freeman's claims inextricably involve a reexamination of military decisions beyond its power to conduct. The judgment of the court of appeals is reversed and *261judgment is rendered that all claims be dismissed.
Justice Guzman filed a dissenting opinion.
Justice Devine filed a dissenting opinion, in which Justice Guzman joined.
Over the past two decades, the military's use of private contractors to support its overseas missions has skyrocketed.1 "At times, the number of contract employees has exceeded the number of military personnel alongside whom they work in these warzones."2 In a decision carrying serious ramifications for those injured by private contractors in combat zones, the Court holds that contractors can escape liability for their actions merely by pointing the finger at the military. The Court's analysis turns on a dangerous misapplication of the political question doctrine and runs counter to our plea-to-the-jurisdiction jurisprudence. I therefore join JUSTICE DEVINE 's dissenting opinion and write separately to expound on these substantive and procedural shortcomings.
I
"[T]he Judiciary has a responsibility to decide cases properly before it, even those it would gladly avoid."3 The political question doctrine is a "narrow exception" to that charge,4 applying only when a political question "is inextricable from the case at bar."5 But with virtually no United States Supreme Court guidance on the topic, courts have been inconsistent in determining how entwined a political question must be for it to be "inextricable" from a case. Multiple approaches have been employed, and this case presents a prime example of the lingering uncertainty.
The Court views the Army as a responsible third party on AMK9's mere say so and dismisses the case without any evidence of that fact, concluding that simply designating the Army as a potentially responsible party means the merits of the case could never be determined without evaluating the military's battlefield decisions. I believe courts must first determine whether a fact issue exists that could obviate any need to assess the military's decisions-here, whether the Army actually caused an injury. Other Courts have taken different analytical paths, such as declining to focus the inextricability determination on the defensive theories that have been asserted-as the Court does here-because that " 'give[s] defendants too much power to define the issues.' "6 Though the existing political-question jurisprudence is fairly well-developed, it is decidedly uneven regarding inextricability, and the Supreme Court has not weighed in to settle the matter.
One thing is clear, however; federal courts confronting the issue have applied a much more searching standard than the Court adopts today, defining inextricable to mean the political question is certain,7 *262required,8 and impossible to avoid.9 Declining to dismiss a suit unless a political question meets the inextricability standard preserves access to the courts and fulfills the judiciary's obligation to resolve disputes. But, here, the Court gives short shrift to this crucial precept, summarily concluding a merits-based disposition is beyond judicial ken.10 Rather than ensuring the inextricable presence of a political question, the Court holds dismissal is required if a contractor asserts-without evidence-that the military might be a causal contributor. The Court abjures its responsibility to decide justiciable cases by embracing a legal standard that terminates litigation before any determination has been made that a political question is actually in play.
Though a court must be careful not to exercise jurisdiction it lacks, it must be equally careful not to decline to exercise jurisdiction it has.11 The Court strikes the wrong balance here. The bright-line rule the Court adopts (1) favors tortfeasors over injured parties, (2) ignores the Supreme Court's holding that only inextricable political questions render a matter nonjusticiable, and (3) is repugnant to our plea-to-the-jurisdiction precedent. Applying the appropriate legal standard and following proper procedures may ultimately lead to dismissal of LaTasha Freeman's lawsuit. But if the military had no part in causing Freeman's injury, the political question doctrine does not bar a merits-based disposition.
II
Whether a political question necessarily arises here remains to be seen. Discovery is still in the early stages, and the trial court dismissed the case without even ruling on Freeman's request for causation-related discovery. The Court concludes these circumstances portend nothing of consequence, holding a political question exists based solely on the contractor's allegations. This approach is misguided.
The existence of a political question requires a "discriminating inquiry into the *263precise facts and posture of the particular case" rather than "resolution by any semantic cataloguing."12 The Court contravenes this directive by concluding that-despite the evidentiary void and without regard to the lawsuit's procedural posture-a political question is presented whenever a defendant alleges the military contributed to the harm claimed. The Court's analysis to the contrary notwithstanding, the procedural context is significant.
This case was decided on a plea to the jurisdiction, and under our well-settled procedures, naked allegations are not enough to sustain a jurisdictional plea. As we have explained time and again, when the jurisdictional inquiry and merits intertwine, as they do on the causation issue here, dismissal is improper absent proof that jurisdiction is lacking. Because such a plea invokes a summary-judgment type proceeding, any fact disputes must be resolved by the factfinder.13 The trial court does not, as the Court implies, have discretion to ignore the evidence and "dismiss the case early on."14 Courts have discretion regarding when , not whether , the evidence should be considered.15
The point is illustrated in many of our sovereign-immunity cases. For example, a plea to the jurisdiction alleging sovereign immunity may not be granted on the bare assertion that the governmental unit was not grossly negligent and was thus immune from suit under the Tort Claims Act.16 If, as here, the plea proceedings go beyond the pleadings, the trial court determines whether the evidence creates a fact issue regarding subject-matter jurisdiction, and if so, the case cannot be dismissed.17 Indeed, a governmental entity may not be released from the case until the factfinder has resolved all facts necessary to determine the jurisdictional matter.18 Concerning the political question doctrine, if a fact issue about the Army's responsibility exists, the political question doctrine may-depending on inextricability-preclude judicial resolution of the disputed fact. But in both sovereign-immunity and political-question cases, dismissal is neither warranted nor required unless evidence raising a fact issue is produced in the first instance.
The Court handwaves established procedure, saying fealty to our precedent would reduce the political question doctrine "to an irrelevance."19 But this is little more than unfounded hyperbole. The Court fails to acknowledge that the political-question and sovereign-immunity doctrines involve common concerns, such as the need to avoid judicial second-guessing.20 Our plea-to-the-jurisdiction *264procedures have not rendered sovereign immunity a nullity, and the political question doctrine is no more endangered by those procedures. The Court's analysis is conspicuously bereft of a compelling justification to jettison established precedent in favor of a special rule for political-question cases.
And this is not the only defect in the Court's analysis. A more disconcerting error lies in the evidentiary void the Court downplays. Because discovery was prematurely halted, the facts of this case have not been developed. The Court's disposition is contrary to the approach taken by federal appellate courts, which look to the evidence, not the allegations, to determine whether a political question is genuinely in play. In case after case, federal courts have remanded for additional discovery and other proceedings necessary to determine whether a political question is actually-rather than potentially-inextricable from the case.21
This is a sound course of action we would be wise to follow because it fulfills our obligation to take cases that may be decided without encroaching on matters committed to the political branches and accords with our plea-to-the-jurisdiction procedures. But even though Freeman is entitled to jurisdictional discovery before her case is dismissed, the Court says Chapter 33's proportionate-responsibility provisions make evaluation of the military's role inevitable. On this point, the Court is demonstrably incorrect because a factual determination that the military was involved in the chain of causation is not equivalent to finding the military responsible.22 Even if Chapter 33 inexorably implicates the military's liability, dismissal at this juncture is precipitous because the Army may not remain designated as a responsible third party. Though the trial court must grant a request to designate a responsible third party if the designation is supported by the pleadings and timely requested,23 Chapter 33 requires the court to strike the designation if, after discovery, the proponent cannot back up its claim with evidence:
*265After adequate time for discovery, a party may move to strike the designation of a responsible third party on the ground that there is no evidence that the designated person is responsible for any portion of the claimant's alleged injury or damage. The court shall grant the motion to strike unless a defendant produces sufficient evidence to raise a genuine issue of fact regarding the designated person's responsibility for the claimant's injury or damage.24
Chapter 33 thus allows Freeman to extricate the Army from the case and avoid a political-question dismissal if evidence of responsibility is lacking after an adequate time for discovery. Yet Freeman has been denied the benefit of the safeguards the statute provides. The district court granted AMK9's plea to the jurisdiction a mere five days after granting the motion to designate the Army as a responsible third party. Five days is rarely an adequate time for discovery. And more importantly, AMK9 has not produced any evidence that the Army caused or contributed to Freeman's injury. The Court is affording the Army's designation as a responsible third party far more weight than Chapter 33 allows.
The Court does not mention Section 33.004(l ), merely stating that the trial court has discretion to dismiss cases raising jurisdictional issues early in the proceedings.25 But Chapter 33 does not allow such discretion; it requires the trial court to strike a responsible-third-party designation if the defendant cannot-with evidence-support the designation. This scenario would, in this case, eliminate the existence of any political question.
Under Chapter 33, a responsible-third-party designation is permitted only on the terms and conditions provided in that statute, and through the process provided in Section 33.004(l ), the alleged political question may be extricated from this case.26 As the adage goes, he who lives by the sword today, may die by the sword tomorrow. Thus, rather than summarily dismissing on the basis of a political question, the cause should be remanded to the trial court to allow the parties to more fully engage the discovery process as required by Chapter 33, our plea-to-the-jurisdiction precedent, and the inextricability requirement.
III
The Supreme Court's most recent political-question guidance serves as a reminder that courts must not shirk their "responsibility to decide cases properly before [them]."27 In Zivotofsky v. Clinton , the Court refused to find a political question precluded the third branch from passing on the constitutionality of certain parts of the Foreign Relations Act.28 In so holding, the Court emphasized the judicial branch's duty to decide cases, observing that judges "appropriately exercise[ ]" the authority to determine the constitutionality of statutes on a regular basis.29 "This is what courts do."30
*266As a concurring opinion in Zivotofsky put it, courts may not "decline to resolve a controversy within their traditional competence and proper jurisdiction simply because the question is difficult, the consequences weighty, or the potential real for conflict with the policy preferences of the political branches."31 After all, deciding such cases "is the role assigned to courts by the Constitution."32
Here, the Court abdicates that role in favor of a bright-line rule that unnecessarily and improperly tilts to the advantage of tortfeasors, allowing wrongdoers to evade responsibility by accusing others. Dismissal on "the mere chance that a political question may eventually present itself" is inappropriate33 and works an injustice on those who risk their lives working alongside military contractors. Because we do not know now, with any certainty, that the potential political question cannot be extricated from this case, dismissal is premature and improper. I respectfully dissent.
Standards of review dictate appellate review. The standard here is extremely deferential to LaTasha Freeman, the nonmovant: we view the facts and pleadings in the light most favorable to her and must deny American K-9 Detection Services, LLC's (AMK9's) plea if a fact question about jurisdiction exists that also implicates the case's merits. See Tex. Dep't of Parks and Wildlife v. Miranda , 133 S.W.3d 217, 226-28 (Tex. 2004). This includes when jurisdiction depends on a fact question about proximate cause. Ryder Integrated Logistics, Inc. v. Fayette Cty. , 453 S.W.3d 922, 929 (Tex. 2015) (denying plea to the jurisdiction when the allegations "generate[d] a fact issue" about "proximate cause"). AMK9's plea is based on its allegation that the Army at least partly caused Freeman's injuries; but Freeman alleges that AMK9, not the Army, proximately caused her injuries. This is a classic fact-question.
So long as this fact question remains, we cannot grant AMK9's plea. Yet the Court flips the standard of review on its head by viewing the evidence in the light most favorable to AMK9, the movant. The Court does this through heavy reliance on the pronouncements-some of which are dicta-of several federal courts. I am unconvinced by their reasoning. The U.S. Supreme Court has not endorsed their views on the political-question doctrine in proportionate-responsibility systems, and we are not otherwise bound by their holdings. I would instead hold that when a political-question doctrine claim depends on a causal finding, we cannot dismiss the suit while causation is disputed. Because the Court's dismissal contravenes well-established plea-to-the-jurisdiction jurisprudence, and because no other ground AMK9 or Hill Country Dog Center, LLC asserts can sustain the plea, I dissent.
I. Jurisdiction over Freeman's claims against American K-9 Detection Services, LLC
A. Political-Question Doctrine
AMK9 argues in its plea that we lack jurisdiction because the Army at least *267partly caused Freeman's injuries, thereby implicating the political-question doctrine. A plea to the jurisdiction is a dilatory plea that defeats a cause of action whether the claims have merit or not. Bland Indep. Sch. Dist. v. Blue , 34 S.W.3d 547, 554 (Tex. 2000). A defendant can use a plea to challenge jurisdiction based on the sufficiency of the plaintiff's pleadings or on the existence of jurisdictional facts. Miranda , 133 S.W.3d at 226. AMK9 challenges the jurisdictional facts. "[W]hether undisputed evidence of jurisdictional facts establishes a trial court's jurisdiction is ... a question of law" that we review de novo. Id. ; see also Tex. Nat'l Res. Conservation Comm'n v. IT-Davy , 74 S.W.3d 849, 855 (Tex. 2002). When the evidence is undisputed or does not raise a fact question about jurisdiction, the trial court rules on the plea as a matter of law. Miranda , 133 S.W.3d at 228. But when the evidence (1) creates a fact question about jurisdiction and (2) implicates the case's merits, "the trial court cannot grant the plea" and the fact-finder must resolve the fact question. Id. at 227-28. In such situations, we determine whether a fact question exists by taking all evidence favorable to the nonmovant as true, indulging every reasonable inference and resolving any doubts in her favor. Id. at 228.
This review essentially mirrors our summary-judgment standard: after the defendant presents evidence that the trial court lacks jurisdiction-and when such evidence also implicates the case's merits-the plaintiff must show only that a jurisdictional fact is disputed to survive the plea. Id. This standard of review saves plaintiffs from having to "put on their case simply to establish jurisdiction" in response to a dilatory plea, which "should be decided without delving into the [case's] merits." Bland , 34 S.W.3d at 554. Otherwise, a plaintiff like Freeman "would be required to try [her] entire case" just to show jurisdiction. Id.
Here, Freeman alleges that AMK9 was negligent for leaving Kallie unattended, not properly training her or her handler, not keeping her under restraint, and not securing the kennel. Freeman does not allege that the Army or its kennel design caused her injuries. AMK9 alleges these things. Thus, Freeman's allegations dispute that the Army proximately caused her injuries. This proximate-cause issue is what potentially raises a political question because if the Army caused Freeman's injuries, we might have to evaluate the Army's military decisions as a responsible third-party. The political-question doctrine, however, bars this suit if and only if a political question-here, the Army's military decisions-is "inextricable from the case." Baker v. Carr , 369 U.S. 186, 217, 82 S.Ct. 691, 7 L.Ed.2d 663 (1962).
The elements of proximate cause are cause-in-fact and foreseeability. W. Invs., Inc. v. Urena , 162 S.W.3d 547, 551 (Tex. 2005). "Because proximate cause is ultimately a question for a fact-finder," we must sustain AMK9's plea if Freeman's petition " 'creates a fact question' regarding the causal relationship between [the Army's conduct] and the alleged injuries." Ryder , 453 S.W.3d at 929 (quoting Miranda , 133 S.W.3d at 228 ). AMK9 alleges that it had to use the Army's kennels, and that the Army's failure to place a top on these kennels at least partly caused Kallie's escape. Maybe so. But a juror might reasonably infer that, had AMK9 closed every kennel door, as well as the kennel building's outer door, the lack of a top would have been causally irrelevant.
Even if the Army was aware that this design might allow a dog to scale the internal dividers between kennel pens, that a successful escape was foreseeable to the Army is far from clear. As Freeman points *268out, the Army required AMK9 to close all of the kennel's doors and the kennel was inside a building. Had AMK9 closed either the kennel's or building's doors-as the Army required it to do-Kallie's escape attempt would have been futile. That arguably makes the foreseeability of her escape doubtful. And we resolve doubts in Freeman's favor. See Miranda , 133 S.W.3d at 228.
Furthermore, a juror could reasonably infer that an escape was as foreseeable to AMK9 as to the Army. In fact, because AMK9, not the Army, trained and handled Kallie, a juror could reasonably infer that AMK9 knew better than anyone whether Kallie might escape as she did. By indulging every reasonable inference and resolving any doubts in Freeman's favor, a juror could find that Kallie's escape was not foreseeable to the Army and that AMK9's actions were the escape's cause-in-fact.
But no matter who proximately caused Kallie's escape, a juror could reasonably conclude that AMK9's allegedly negligent training was the attack's sole proximate cause. Freeman alleges that AMK9 trained Kallie such that she-contrary to the Army's contractual specifications-attacked without cause and without being ordered. Viewing these facts favorably to Freeman, a juror could reasonably conclude that a dog trained to the Army's specifications presents no attack threat to people like her. Thus, even if the Army partly caused Kallie's escape, a juror could reasonably conclude that an attack was not foreseeable to the Army and that AMK9's training was the attack's cause-in-fact. We must, therefore, conclude that a fact question about proximate cause exists.
Indulging every reasonable inference and resolving any doubts in Freeman's favor and taking all evidence favorable to her as true, a juror could reasonably conclude that AMK9's actions-and only AMK9's actions-caused the alleged attack. See id. at 228. AMK9, of course, disputes this. It alleges that the Army at least partly caused Kallie's escape and, therefore, this suit necessarily requires evaluating sensitive military decisions. We, however, cannot decide that issue while an underlying causal fact-question exists because that question affects whether the military's decisions are "inextricable from the case." Baker , 369 U.S. at 217, 82 S.Ct. 691. Thus, answering the political-question doctrine issue now, as the Court does, is premature.
The Court actually agrees that causation is a disputed fact-question here that can be decided only by a jury. Ante at 259 (citing Miranda , 133 S.W.3d at 227-28 ). Nevertheless, the Court is unmoved by that bar to its judgment. Despite that we cannot determine whether the doctrine is implicated without first resolving that fact question, the Court retorts that my analysis here, which simply applies our plea-to-the-jurisdiction standard, makes the doctrine "an irrelevance." Ante at 259. That is not true. If AMK9 is correct, it might get the suit dismissed under the doctrine, which is very relevant. We would not reduce the doctrine to an irrelevance by making AMK9 actually prove the facts of its defense. That makes the doctrine no more irrelevant than in any other suit where a potentially dispositive defense depends on a fact question that can be determined only through trial.
Furthermore, the Court's holding has a worrisome consequence to our jurisprudence. The holding essentially bars all tort suits where a military contractor-or any other defendant-is able to muster a mere allegation that a government actor whose decisions are insulated by the political-question doctrine partly caused the alleged harm. Even if the Court's view of that doctrine is otherwise right, its application *269here throws out cases where unproven, disputed factual allegations affect whether the doctrine is, in fact, implicated. That, in my view, throws the baby out with the bathwater.
For example, if a soldier sued a contractor for negligently making a tank hatch contrary to Army requirements such that it did not open properly, trapping him inside and injuring him, the contractor could obtain dismissal by merely alleging that the Army at least partly caused the hatch's failure. The Army, the contractor might argue, decided to park the tank in an area without a cover, and this exposure to the elements caused the hatch to fail. Thus, the Army's wartime military decision partly caused the injury, implicating the political-question doctrine. If that truly did cause the hatch to fail, the contractor might be entitled to dismissal. But if the soldier disputes that the Army's actions caused the failure, instead alleging that the contractor's actions are the sole proximate cause, this causal fact-question should allow him to survive a plea to the jurisdiction. Otherwise, we deny all relief even when the soldier's allegations prove true.
The Court attempts to cabin this slippery slope, but in doing so shows why a jury needs to resolve the fact question here. The Court states that had "Kallie bit Freeman while being routinely exercised by her civilian-contractor handler," the attack "would have had nothing to do with the military." Ante at 260. But that does not solve the problem because in that scenario AMK9 has not alleged that the Army partly caused anything. Based on the Court's opinion here, AMK9 would be foolish not to make such an easy-to-manufacture allegation. For example, AMK9 could argue that the Army partly caused this attack by not providing enclosed yards for exercising these working dogs, thereby implicating the Army's equipment decisions. That causal allegation might be without merit, but that's the point-so might AMK9's actual allegation. The problem is that AMK9 is getting this case dismissed as a matter of law based on a disputed fact-question. The Court's scenario does not solve this problem because the scenario does not address this problem. Instead, the Court's scenario avoids the issue: how do we handle cases where a defendant's disputed causal allegation might implicate a military decision given that the allegation might be wrong? If AMK9 is wrong here, this case, too, has "nothing to do with the military." Id. That is why a fact-finder must resolve this fact question.
Because AMK9's jurisdictional plea and Freeman's case on the merits both depend on the same fact question-whether the Army or AMK9 proximately caused Freeman's injury-we should deny AMK9's plea, leaving this fact question for the fact-finder. See Miranda , 133 S.W.3d at 227-28.
The foregoing analysis should be enough to deny AMK9's plea. The Court, however, effectively side-steps this in holding that AMK9's proximate-cause defense would require the jury to impermissibly evaluate the Army's decisions about the kennel's design and construction. Ante at 252. That completely ignores the possibility that the Army might not be a cause at all. That move, however, is consistent with several federal cases. The Court endorses those cases, but their reasoning on that point cannot withstand scrutiny.
In Harris v. Kellogg Brown & Root Services, Inc. , the Third Circuit held that a sole-proximate-cause defense would not implicate the political-question doctrine because that dispute is "simply about who did what." 724 F.3d 458, 473 (3d Cir. 2013). But the court then concluded that, in a proportionate-responsibility system, determining whether the military was a proximate cause (rather than the sole proximate *270cause) would require a court to impermissibly second-guess military decisions. Id. at 474. In re KBR, Inc., Burn Pit Litigation , relying on Harris , also held that a proximate-cause defense does not make a suit nonjusticiable unless the military at least partly caused the plaintiffs' injuries and the suit was in a proportionate-responsibility system. 744 F.3d 326, 340-41 (4th Cir. 2014). The Harris court's explanation for this distinction was that, when determining partial cause, "there is simply no way to determine damages without evaluating military decisions" because the fact-finder "cannot decide the respective degrees of fault" between the military and the contractor "without evaluating the decisions made by each...." Harris , 724 F.3d at 474. But that does not explain the distinction.
Rather, this explanation skips a step. It incorrectly assumes that finding that the military partly caused the injury means finding that the military negligently caused it. The latter might be a political question, but the former is not. Determining "who did what" does not require second-guessing any decisions, military or otherwise. See id. at 473. Causal questions are objective, not normative. Objective questions do not inexplicably become normative just because Texas uses a proportionate-responsibility system. Holding otherwise conflates the distinction between causation and negligence.
Negligence assessments require multiple findings. In Texas, a court must find (1) the existence of a legal duty, (2) a breach of that duty, and (3) damages proximately caused by that breach. IHS Cedars Treatment Ctr. of DeSoto, Tex., Inc. v. Mason , 143 S.W.3d 794, 798 (Tex. 2004). No party is ever negligent merely by causing an event; something more is always required. This is true as a matter of basic tort law. See RESTATEMENT (SECOND) OF TORTS § 281 (1965) (Statement of the Elements of a Cause of Action for Negligence). A fact-finder also must determine that the party should have acted otherwise. This determination-whether the party should have acted as it did-is the potential political question. "[W]ho did what" is not. Harris , 724 F.3d at 473.
Nothing about our proportionate-responsibility system changes that a causal finding does not second-guess anything. And neither the Court nor its cited authorities explain how it could. Thus, resolving the factual dispute here-whether the Army or AMK9 proximately caused Freeman's injuries-does not raise a political question even if this suit eventually raises one because AMK9's allegations prove true. AMK9's allegations have to actually be true first. This is why I cannot condone the Court's reasoning or its reliance on any case that holds that our proportionate-responsibility system somehow transforms a causal finding into a political question. The causal finding might raise a political question later in the suit, but the finding is not itself a political question.
We cannot avoid a fact question now just because a dispositive legal one might arise later. At no point in a factual who-did-what determination will the court or jury re-examine a military decision. See id. Why the Army made that decision and whether the decision was justified are irrelevant to that inquiry. The decision was made. All that matters is whether it caused Freeman's injuries, as AMK9 claims, or not, as Freeman claims.
This untenable distinction between causation defenses is rendered even more inexplicable when we consider that, had AMK9 argued only that the Army was the sole proximate cause, the Court would not dismiss this suit-that defense does not raise a nonjusticiable issue. Ante at 256 (citing Harris , 724 F.3d at 473 ;
*271In re KBR , 744 F.3d at 340-41 ). Apparently, a defendant is better off admitting that he partly caused an injury than that he did not cause it at all, so long as he also asserts that the military partly caused the injury-i.e. , admitting partial fault will get your case dismissed; denying fault completely will not. That cannot be right, and is probably why the U.S. Supreme Court has not endorsed this view.
The Court dismisses this entire suit on the mere allegation that the Army might have at least partly caused Freeman's injuries. Because that causal fact-question is disputed, and because nothing about our proportionate-responsibility system mutates such a causal finding into a political question, we should not yet hold that a political question is "inextricable from the case." Baker , 369 U.S. at 217, 82 S.Ct. 691. Until a political question is so intertwined, I cannot join the Court's judgment.
B. Preemption under the Federal Tort Claims Act
AMK9 argues that the Federal Tort Claims Act preempts Freeman's Texas tort-law claims because the Act's combatant-activities exception applies here. See 28 U.S.C. § 2680(j) (2012). The Tort Claims Act is a limited waiver of sovereign immunity, 28 U.S.C. § 2674, and under the combatant-activities exception, the United States retains its immunity for "[a]ny claim arising out of the combatant activities of the military or naval forces, or the Coast Guard, during time of war." Id. But "[u]nlike complete preemption, which is a jurisdictional issue," preemption based on the combatant-activities exception is "only an affirmative defense." McManaway v. KBR, Inc. , 852 F.3d 444, 447 n.2 (5th Cir. 2017) ; see also Spear Mktg., Inc. v. BancorpSouth Bank , 844 F.3d 464, 467 n.3 (5th Cir. 2016), Cmty. State Bank v. Strong , 651 F.3d 1241, 1260 n.16 (11th Cir. 2011). "[N]o court has held[ ] that" preemption under this exception "constitutes complete preemption," and "[a]bsent complete preemption, whether a plaintiff's claims are preempted relates to the merits." Harris , 724 F.3d at 463. I see no reason to disagree with the federal circuits on this matter.
The combatant-activities exception does not preempt all state-law tort claims; it preempts only those claims "arising out of" combatant activities. 28 U.S.C. § 2680(j). Such claims are only a small subset of potential tort claims, not the entire substantive field of tort claims or even the entire field of tort claims against contractors overseas. See Spear Mktg. , 844 F.3d at 467 n.3. Because the combatant-activities exception does not convert all such state-law tort claims into federal claims-i.e. , despite the exception, state-law tort claims continue to exist-the exception is merely "ordinary" preemption, not "complete" preemption. See GlobeRanger Corp. v. Software AG , 691 F.3d 702, 705 (5th Cir. 2012) (quoting Caterpillar Inc. v. Williams , 482 U.S. 386, 393, 107 S.Ct. 2425, 96 L.Ed.2d 318 (1987) ); 14B CHARLES ALAN WRIGHT ET AL. , FEDERAL PRACTICE & PROCEDURE § 3722.2 (4th ed. 2016). AMK9's preemption argument is, therefore, only a defense. Even if the exception applies to Freeman's claims, the exception goes to the merits and, hence, cannot sustain a jurisdictional plea.
C. Westfall Immunity
AMK9 next argues that it is entitled to Westfall immunity, a form of absolute official immunity. This argument was not one of AMK9's original bases for its jurisdictional plea. AMK9 originally argued derivative sovereign immunity, but on appeal has abandoned that ground in favor of Westfall immunity. Because Westfall immunity is immunity from suit, defendants *272can raise it for the first time on appeal. San Antonio Water Sys. v. Nicholas , 461 S.W.3d 131, 136 (Tex. 2015).
The original test for determining whether absolute official immunity applies comes from Westfall v. Erwin , 484 U.S. 292, 108 S.Ct. 580, 98 L.Ed.2d 619 (1988). That decision has been superseded by statute, see 28 U.S.C. § 2679(d) (2012), but the Westfall test is still used to determine when such immunity applies to nongovernmental entities for state-law tort claims. Houston Cmty. Hosp. v. Blue Cross and Blue Shield of Texas, Inc. , 481 F.3d 265, 269 (5th Cir. 2007); accord Murray v. Northrop Grumman Info. Tech., Inc. , 444 F.3d 169, 174 (2d Cir. 2006) ; see also Beebe v. Washington Metro. Area Transit Auth. , 129 F.3d 1283, 1289 (D.C. Cir. 1997). Under this test, a nongovernmental entity is entitled to immunity when it makes discretionary decisions within the scope of its duties to perform an official government function. Houston Cmty. Hosp. , 481 F.3d at 269.
Here, AMK9 handles bomb-sniffing working dogs at a forward-operating base during war. These dogs work with active combat-units in the field, supplementing the military's own working dogs. Such overseas combat-related work that is integrated with the military is quintessentially governmental in nature. But Freeman is not alleging that AMK9 erred in any of its discretionary acts while performing these government functions. Freeman is claiming that AMK9 failed to do what the Army required-i.e. , to close all of the kennel's doors, to not leave any dogs unattended, and to train the dogs so that they would attack only when ordered or given cause. AMK9 did not have discretion to violate its contractual duties or the Army's policies. Indeed, official immunity is not meant "to protect an erring official, but to insulate the decision-making process" from litigation. Westfall , 484 U.S. at 295, 108 S.Ct. 580. The Army already prescribed AMK9's actions-the relevant official decisions were already made. Hence, Freeman is not challenging AMK9's discretionary decisions; she is challenging its failure to do what the Army already decided that AMK9 must do. Whether these alleged failures proximately caused Freeman's injuries is a separate question that goes to the merits. But for purposes of evaluating AMK9's immunity argument, AMK9 has failed to demonstrate that it is immune from suit by, as Freeman alleges, not doing what the Army required.
AMK9 points out, however, that it had discretion in how to train its dogs. True, but Freeman is not challenging AMK9's discretionary decisions in picking particular training methods. She claims that AMK9 failed to deliver working dogs that met the Army's performance-based contract requirements. Performance-based contracts "describe the work in terms of the required results rather than ... 'how' the work is to be accomplished...." Saleh v. Titan Corp. , 580 F.3d 1, 10 (D.C. Cir. 2009) (quoting 48 C.F.R. § 37.602(b)(1) ). One of those requirements was that these dogs would attack only when commanded or when given cause. AMK9 therefore had discretion in how to train its dogs to meet these requirements, not whether its dogs met them. For that reason, this argument fails. Thus, Westfall immunity cannot sustain AMK9's plea to the jurisdiction.
D. Defense Production Act
Finally, AMK9 argues that it is immune from suit because its contract with the Army is a "rated order" contract under the Defense Production Act. The Act authorizes the President to "require that performance under contracts or orders ... which he deems necessary or appropriate to promote the national defense shall take *273priority over performance under any other contract or order...." 50 U.S.C. § 4511(a) (Supp. IV 2016). The Act later states that "[n]o person shall be held liable for damages or penalties for any act or failure to act resulting directly or indirectly from compliance with a rule, regulation, or order issued pursuant to this [Act]...." Id. § 4557. Hence, the Act provides immunity to contractors who give their "rated order" contracts priority over other contracts or orders when their actions (or inactions) in doing so might otherwise subject them to liability.
Assuming that the Act applies here, it cannot sustain AMK9's jurisdictional plea. Even though the Act "plainly provides immunity," it does so "[b]y expressly providing a defense to liability." Hercules Inc. v. United States , 516 U.S. 417, 429, 116 S.Ct. 981, 134 L.Ed.2d 47 (1996) (emphasis added). The Act, therefore, provides immunity from liability, not suit. See Brown & Gay Eng'g, Inc. v. Olivares , 461 S.W.3d 117, 121 (Tex. 2015) ("Immunity from liability is an affirmative defense ... while immunity from suit bars suit against the entity altogether and may be raised in a plea to the jurisdiction."). Thus, whether the Act applies to tort suits like this one or not, AMK9's Defense Production Act-based defense cannot sustain its jurisdictional plea.
* * * * *
Because none of AMK9's arguments establish a lack of subject-matter jurisdiction, the trial court erred in granting AMK9's plea to the jurisdiction.
II. Jurisdiction over Freeman's claims against Hill Country Dog Center, LLC
Hill Country Dog Center did not file a plea to the jurisdiction. It did file a Rule 91(a) motion, but the trial court did not rule on it. See TEX. R. CIV. P. 91(a). On appeal, Hill Country argues that, under Texas law, liability for a dog attack runs only to the owner at the time of the incident, not to the former owner. It also argues that no causes of action for negligently training a dog or for strict liability for a non-owner exist. Hence, Hill Country argues that the trial court was correct to find, sua sponte , that it lacked jurisdiction over Freeman's claims.
Whether these arguments are correct statements of Texas law or not, Hill Country provides no authority that they deprive the trial court of subject-matter jurisdiction. Its arguments, even if meritorious, offer immunity only from liability, not suit. Thus, the trial court erred in dismissing Freeman's claims against Hill Country based on a lack of subject-matter jurisdiction.
III. Conclusion
Freeman alleges that AMK9's supervision and training of its dog was the cause-in-fact of her injuries. AMK9 alleges that the Army was partly to blame. Although the Court does not know whether either allegation is true, it nonetheless dismisses Freeman's claim against AMK9 because the Army might have contributed to causing her injuries. Even assuming that the Court otherwise correctly applies the political-question doctrine to such partial-cause scenarios, I simply cannot understand how the mere allegation that the Army might have partly caused Freeman's injuries is sufficient to defeat her claim-a claim that does not even raise that issue. The Court ignores these deficiencies to sustain AMK9's jurisdictional plea notwithstanding the existence of unresolved factual questions necessary to the doctrine's application. Thus, I respectfully dissent.