dissenting:
The facts of this case are tragic. A boy lost his life—needlessly, as it turns out. We know now that he was carrying only a fake gun, albeit a realistic-looking one. Deputies Gelhaus and Schemmel therefore never were in any real danger and deadly force was not necessary. In view of these facts, the inclination to hold Deputy Gel-haus liable for shooting Andy Lopez is understandable. But it is a well-settled rule that a court may do so only if precedent clearly established at the time of the shooting that the use of deadly force in the circumstances Deputy Gelhaus faced was objectively unreasonable. I do not agree with the majority that such a case existed on the day Andy died. Respectfully, I therefore dissent.
I.
The majority opinion exhaustively recounts the facts of the case, but for me, they are largely irrelevant. One critical fact—the upward motion of the fake gun— resolves the qualified immunity issue in Deputy Gelhaus’s favor. In reaching the opposite conclusion, the majority accuses me of making an assumption regarding this fact that is improper at the summary judgment stage. I have done no such thing. In fact, as I explain below, it is the majority whose position is unsupported by the record. For contextual purposes, and to rebut any contrary implication in the majority opinion, I also will explain why the statements of Jose Licea, who testified regarding the appearance of Andy’s fake gun, do not affect the qualified immunity analysis.
A.
As the majority concedes, we must accept the district court’s finding that the barrel of the gun “was beginning to rise.” The majority also accepts the district court’s additional finding that the gun “could have been raised to a slightly-higher level without posing any threat to the officers.” Based on the latter finding, I agree with the majority that we must assume the gun was not in fact pointed at the officers at the moment Deputy Gelhaus opened fire. As the' majority says, neither Deputy Gelhaus nor Deputy Schemmel testified how high the gun barrel rose, but both stated that they believed they were in imminent danger as a result of the gun’s movement. This evidence shows that the deputies at least perceived that the weapon posed a threat at the height to which it had then risen. Their perception is not dispositive, however, and there is other evidence in the record (and the district court’s finding) that the gun, while rising, had not yet risen to a point where it could have shot either deputy. I agree with the majority, therefore, that the precise angle at which Andy pointed the gun is a disputed fact, but as I explain below, that fact is not material to the qualified immunity analysis.
The majority attempts to discount the district court’s finding that the gun barrel was beginning to rise. For instance, in summarizing the facts in the light most favorable to the plaintiffs, the majority says that “[Deputy] Gelhaus deployed deadly force while Andy was merely standing on the sidewalk holding a gun that was pointed down at the ground.” This description does not characterize fairly the situation that Deputy Gelhaus faced. A gun pointed at the ground and one that is rising are qualitatively different. By casting the latter as the former, the majority goes beyond viewing the facts in the light most favorable to the plaintiffs and ignores a critical fact that must be accepted as true and, as I will explain, bears directly on the question of whether it was clearly established that Deputy Gelhaus’s use of deadly force was unreasonable under the circumstances. The majority repeats this error when it describes the record as showing that “as Andy engaged in the turn, the position of the gun barrel never posed any threat to [Deputy] Gelhaus” without a mention of the gun’s upward motion.
The majority takes me to task for “rel[ying] on the assumption that Andy’s gun was continuously rising throughout the interaction,” an assumption that the majority believes is unsupported by the record. This criticism is puzzling for two reasons. First, I have not taken Deputy Gelhaus’s “word at face value,” as the majority charges. What I have done, and I was under the impression that the majority had done the same, is accept the district court’s finding that the fake gun’s barrel “was beginning to rise.” Estate of Lopez v. Gelhaus, 149 F.Supp.3d 1154, 1162 (N.D. Cal. 2016). Not only is it not improper for me to accept this fact, it is required. Watkins v. City of Oakland, 145 F.3d 1087, 1091 (9th Cir. 1998). The majority itself embraces this finding as one that “makes sense.” So, as far as I can tell, the majority’s concern is one of timing'—that although the barrel may have begun to rise at some point before the shooting, it may also have ceased to rise in time for Deputy Gelhaus to recognize that Andy did not pose a threat.
This position is difficult to reconcile with the district court’s finding. The district court did not find that the gun’s barrel stopped moving after beginning to rise. It found only that the barrel “was beginning to rise.” Lopez, 149 F.Supp.3d at 1162. To the extent the majority believes some ambiguity exists as to whether the district court found that the gun was still rising immediately before Deputy Gelhaus shot Andy, the court’s legal analysis confirms my reading. It found that the gun “was beginning to rise” while distinguishing cases involving shootings preceded by actions that, from the district court’s perspective, were more threatening. Id. If the district court wanted to distinguish those cases on the basis that Andy’s action was not sufficiently threatening, it would make little sense to find that the gun barrel was “beginning to rise” if there was room to find instead that the gun barrel had stopped rising. Therefore, the most natural reading of the district court’s finding, and the only reasonable one, is that the gun was beginning to rise (i.e., in the process of rising) immediately before Deputy Gelhaus shot Andy.
This brings us to the second flaw in the majority’s argument, which is that it is completely unsupported by the record. The majority speculates that the gun may not have been rising at the time Deputy Gelhaus committed to firing his weapon. Contrary to the majority’s contention, however, nothing in the record before us supports this proposition. The majority’s reliance on the three-dimensional models created by the plaintiffs’ expert is misplaced. Those models are components of the expert’s analysis of Andy’s likely body posture at the time he was struck by the bullets. As such, they necessarily concern only what occurred after Deputy Gelhaus first fired his weapon and thus cannot serve as evidence of the gun’s motion even at the moment of the shooting, much less at the time Deputy Gelhaus became committed to using deadly force. With respect to Deputy Gelhaus’s purported admission “that the gun had been benignly swinging ... with Andy’s natural motions,” it is true that Deputy Gelhaus stated that none of Andy’s “motions” during the time leading up to the confrontation seemed aggressive, and that the gun would “swing somewhat” as Andy walked. Even granting that the gun was moving in this way while Andy was walking away from the police car, however, that fact does not tell us how the gun moved when Andy stopped walking and engaged in an entirely different motion—namely, turning to face Deputy Gel-haus.
The majority has thus identified no evidence that even suggests that the gun had stopped rising at the time Deputy Gelhaus resorted to deadly force. This dearth of support might explain why the plaintiffs themselves have never made such an argument, preferring instead to contest whether the gun began to rise at all. Even the majority seems to recognize that the evi-dentiary foundation for its argument is lacking, as it does not claim that the evidence just discussed in fact supports a finding that the gun stopped rising. Instead, the majority asserts only that this evidence is reason to doubt my “assumption” (which really is nothing more than a reasonable, natural reading of the district court’s finding) in the abstract. To reach its ultimate conclusion, the majority cites Deputy Gelhaus’s statement that he did not know where Andy’s gun was pointing when he pulled the trigger and declares that the gun “did not necessarily rise throughout the whole interaction.” At bottom, then, the majority’s argument rests on the bare absence of evidence definitively disproving the existence of alternate facts for which there is no record. My “seismic” “accusations,” as the majority calls them, are a straightforward reading of the district court’s finding.
This novel rule—that we must accept as true all facts not conclusively disproved by evidence in the record even if those facts have no evidentiary support of their own— is plainly wrong. We need only “assume the truth of the evidence set forth by the nonmoving party with respect to [a] fact” when “direct evidence produced by the moving party conflicts with direct evidence produced by the nonmoving party.” T.W. Elec. Serv., Inc. v. Pac. Elec. Contractors Ass’n, 809 F.2d 626, 631 (9th Cir. 1987) (emphasis added). The record before us contains evidence (not to mention the district court’s finding) that the gun was beginning to rise, but no evidence showing that the gun then stopped rising before the shooting started. The mere possibility that a jury might disbelieve a moving party’s undisputed evidence is not enough to avoid summary judgment. See id. at 630 (“[T]he nonmoving party may not merely state that it will discredit the moving party’s evidence at trial and proceed in the hope that something can be developed at trial in the way of evidence to support its claim”). What the majority has done here is to conjure up “some metaphysical doubt as to the material facts”—a step that not even the district court took—and affirm the denial of summary judgment on that basis. Matsushita Elec. Indus. Co., Ltd. v. Zenith Radio Corp., 475 U.S. 574, 586, 106 S.Ct. 1348, 89 L.Ed.2d 538 (1986).
This attempt to avoid the conclusion that the gun was rising at the time Deputy Gelhaus decided to use deadly force is unpersuasive, but also unsurprising. As discussed below, none of the cases the majority cites to show that Deputy Gel-haus violated Andy’s clearly established right addressed a situation where the victim’s gun “was beginning to rise” toward the officer. So to make those cases fit, the majority must eliminate this crucial differentiating fact. Perhaps knowing that the district court’s finding of fact cannot so casually be cast aside, the majority unper-suasively attempts to parse the district court’s language to create a distinction between the gun’s initial motion and its continuing motion and concludes that the district court’s finding pertains only to the first. This clever argument leaves the majority free to attribute any conclusion about the second to some other source— here, Deputy Gelhaus’s account—and then chide me for misunderstanding the rules of summary judgment. Unfortunately for the majority, nothing in the district court’s order reflects that it even considered this dichotomy, let alone structured its factual findings around it.
But the majority does not stop there. It then subtly revises the district court’s finding to make it appear consistent with this new reading. According to the majority, the district court found that “the barrel of the weapon could incidentally have risen, as part of the natural turning motion, only ‘to a slightly-higher level [that did not] pos[e] any threats to the officers.’ ” Tellingly, the majority inserts “only” here in nearly every place (seven to be exact) that it purports to quote or paraphrase this finding. This seemingly innocuous insertion greatly distorts the finding in a way that supports the majority’s argument. On this reading, the district court made a finding that, at the time Deputy Gelhaus committed to using deadly force, the gun could not have risen to a level where it threatened the officers. Had the district court made this finding, it would indeed support the majority’s argument. But this alteration is important. What the district court actually found was that “the rifle barrel was beginning to rise; and given that it started in a position where it was pointed down at the ground, it could have been raised to a slightly-higher level without posing any threat to the officers.” Lopez, 149 F.Supp.3d at 1162. This language paints a different picture: far from concluding that the gun could not rise further, the district court found that the rising motion was not necessarily sufficient to put the gun in a position where it was pointed at the officers. If anything, then, the language that the district court actually used reinforces the notion that it found that the gun was moving when Deputy Gelhaus decided to fire his weapon. In addition, even under the majority’s distorted reading, the gun was necessarily pointed somewhere between the ground and Deputy Gelhaus, by virtue of it “beginning to rise” after having been “pointed down at the ground.” Therefore, the gun was not “trained on the ground” or “pointed down at the ground” at the time Deputy Gelhaus pulled the trigger as the majority claims.
The majority says it is deferring to the district court’s findings, but it is not. Rather than perform these interpretive changes, I would take the district court at its word and decide this appeal on the understanding that the gun was beginning to rise when Deputy Gelhaus committed to using deadly force.
B.
The majority also fails to appreciate the apparent threat posed by the gun from Deputy Gelhaus’s perspective. The record is replete with evidence that Deputy Gel-haus did not realize and could not have discerned that Andy was carrying a fake gun instead of an authentic AK-47. First, it is undisputed that the gun was missing the bright orange tip required by federal law. 15 U.S.C. § 5001(b)(1). This.tip immediately would have identified the gun as a fake; conversely, its absence would suggest to an observer that the gun was real.
Second, Deputy Gelhaus, who had experience with AK-47s both as a deputy and during his time serving in the United States Army, testified that he believed Andy was carrying a real AK-47 and that “[tjhere were no unusual markings or colorings on the weapon which were visible to [him] which indicated that the weapon was anything other than an AK[-]47 assault weapon.” Furthermore, he testified that it was not until after the shooting, when he was close to the gun, that he was able to recognize that it was not a real rifle.
We cannot simply take Deputy Gel-haus’s word, however. As the majority counsels, we must instead “carefully examine all the evidence in the record ... to determine whether the officer’s story is internally consistent and consistent with other known facts.” Cruz v. City of Anaheim, 765 F.3d 1076, 1079 (9th Cir. 2014), quoting Scott v. Henrich, 39 F.3d 912, 915 (9th Cir. 1994). Here, there is expert testimony that it was not possible for Deputy Gelhaus to tell the difference between Andy’s fake gun and a real AK-47 at the distances from which he observed it. Plaintiffs’ expert does not refute this conclusion, opining only that the reenactment video upon which Deputy Gelhaus’s expért relied “does not necessarily accurately depict the information available to Deputy Gelhaus prior to his decision to shoot [Andy].” Notably absent is any direct evidence that a reasonable officer in Deputy Gelhaus’s position would have been able to differentiate between the fake and the real thing.
The majority’s factual exposition refers to statements by Jpse Licea, a witness who observed Andy walking on the sidewalk before the shooting. Licea testified that, as he drove by Andy, he thought the gun “look[ed] fake.” Taken at face value, his assessment of the fake gun’s appearance might seem to create a genuine dispute of material fact. A bit of digging, however, reveals that Licea’s perception was based largely on facts and circumstances unique to him. For example, he qualified his statement that he thought Andy was carrying a BB gun by explaining that someone had recently shot a window at his house, after which his mother-in-law observed some children with BB guns in the area. There was no reason for Deputy Gelhaus to know this fact, so it should play no part in the analysis. White v. Pauly, — U.S. —, 137 S.Ct. 548, 550, 196 L.Ed.2d 463 (2017) (per curiam) (“Because this case concerns the defense of qualified immunity, ... the Court considers only the facts that were knowable to the defendant'officers” (emphasis added)).
.Licea’s perception of the fake gun was also influenced by his assumption that no one would be carrying an AK-47 during the daytime. “[Tjhat’s something for the night,” he asserted. Putting aside the reasonableness of this assumption as. a general matter, it is not one that a reasonable officer in Deputy Gelhaus’s position would have shared, given that the area had a very high concentration of weapons-relatéd violent crime and Deputy Gelhaus himself previously had confiscated an authentic AK-47 within a mile of the site of the shooting. Licea’s assumption therefore should be discounted as well.
Other than these two personal reasons, Licea offered no basis for his conclusion that the gun appeared not to be real. Most importantly, he did not identify anything about the gun itself that gave him that impression. The only time he mentioned the gun’s appearance—which is the only information that was available to Deputy Gelhaus—was a remark that its shape and design, particularly with respect to the clip, “made it look like an AK-47.” Because Licea’s opinion that the gun looked fake is grounded not in objective facts, but rather in his own idiosyncratic understandings, it does not create a genuine dispute of material fact with respect to whether a reasonable officer in Deputy Gelhaus’s position would have been able to distinguish between Andy’s replica and a real AK-47.
This conclusion is not inconsistent with our limited role in reviewing the denial of qualified immunity on summary judgment. It is true that “[a]ny decision by the district court ‘that the parties’ evidence presents genuine issues of material fact is categorically unreviewable on interlocutory appeal.’ ” George v. Morris, 736 F.3d 829, 834 (9th Cir. 2013), quoting Eng v. Cooley, 552 F.3d 1062, 1067 (9th Cir. 2009). In fact, the district court made no such decision with respect to Deputy Gelhaus’s ability to discern that the gun was not real. Instead, the district court stated first that it was setting that issue aside, Lopez, 149 F.Supp.3d at 1158 n.l, and then later that “even assuming the reasonableness of that belief [that the fake was a real gun], qualified immunity is still not warranted.” Id. at 1164 n.2. Nowhere did the district court say that there was a genuine dispute of material fact regarding the possibility of identifying the gun as a fake. We therefore are not constrained in our analysis of that issue.
If anything, the district court’s assumption that Deputy Gelhaus’s perception was reasonable points in the other direction. Where there is a genuine dispute of material fact, the “facts must be viewed in the light most favorable to the nonmoving party.” Scott v. Harris, 550 U.S. 372, 380, 127 S.Ct. 1769, 167 L.Ed.2d 686 (2007) (emphasis added). Clearly it would have been more favorable to the plaintiffs if Deputy Gelhaus treasonably perceived the fake gun to be real, so if the district court believed there was a genuine dispute of material fact on that- issue, it was obliged to so construe the facts. We presume that district courts follow the law, United States v. Cervantes-Valenzuela, 931 F.2d 27, 29 (9th Cir. 1991), and nothing in the record before us suggests that the contrary is true in this case. There is therefore no reason to depart from the district court’s decision to assume that Deputy Gelhaus reasonably believed the gun to be real.
In sum, I reject the false dichotomy the majority has created with respect to the movement of the gun. The district court found that the barrel was “beginning to rise” without distinguishing between an initial rising motion and a continuing rising motion. I would adhere to that finding. Furthermore, I emphasize that there is no genuine dispute of material fact as to whether a reasonable officer in Deputy Gelhaus’s position could have recognized that the gun was not real. Finally, as the majority and plaintiffs concede, it is undisputed that Andy failed to drop the gun after officers activated the patrol car lights and siren, and yelled at him at least once to drop the gun. Accepting these facts, I turn to the question of clearly established law.
II.
I agree with the majority’s conclusion that the district court erred by failing to conduct the necessary analysis identifying a precedential case or cases it believed would have put Deputy Gelhaus on notice that his conduct was unconstitutional. White, 137 S.Ct. at 552. Rather than conclude there and decide the appeal, the majority attempts to perform on its own the district court’s task by identifying three cases—not one of which appears anywhere in the district court’s order— that purportedly served as notice to Deputy Gelhaus that he could not constitutionally use deadly force against Andy.
More important than the district court’s omission, which should require reversal, is that the plaintiffs themselves have never argued that these cases clearly established Andy’s right, either in response to Deputy Gelhaus’s motion for summary judgment or in their answering brief on appeal. As the majority recognizes, “[t]he plaintiff bears the burden of showing that the right at issue was clearly established under this second prong” of the qualified immunity analysis. Sorrels v. McKee, 290 F.3d 965, 969 (9th Cir. 2002). The majority’s effort improperly attempts to carry plaintiffs’ burden for them. This is yet another reason to reverse the district court.
In addition to contravening settled law, the majority’s defense of the district court’s incomplete holding is ultimately unsuccessful on the merits. In my view, all of the eases cited are distinguishable on their facts from the one before us and therefore cannot perform the function the majority ascribes to them, even if it were appropriate for the majority to attempt to do so.
A.
The majority relies primarily on our case of George v. Morris. In that case, the defendant officer shot the victim, who was armed with a pistol and had been reported as acting erratically, after he allegedly “grasped the gun with both hands” and pointed it “directly at” the officer. 736 F.3d at 833 n.4. We could not credit the officer’s account, however, because the district court had found it to be disputed. Id. Importantly, there was evidence in the record that “called into question whether [the victim] ever manipulated the gun.” Id. at 833 (emphasis added). In the most favorable light, then, the victim did not manipulate the gun before the officer resorted to deadly force. See id. at 839 (describing the victim’s gun as “trained on the ground”). This fact conclusively distinguishes George from the case before us because Andy did manipulate the gun—it was beginning to rise toward the deputies as he turned. Here again the majority tries unsuccessfully to evade the district court’s factual finding that the gun “was beginning to rise” so that it can also avoid this manipulation issue. Since the majority is wrong on the first point for the reasons already mentioned, its second point is a non-issue.
Given the version of the facts it was required to assume, the court in George had no occasion to pass judgment on the use of deadly force in a situation like the one Deputy Gelhaus faced. George may have clearly established that using deadly force against an armed individual is unreasonable when that person does not “ever manipulate[] the gun,” id., but that rule says nothing about the use of such force when someone does manipulate a gun. Indeed, our court took pains to emphasize that we were not considering the officer’s version of events, according to which the victim had done just that. Id. at 833 n.4, 838.
The majority’s attempt to shoehorn the facts of our case into George is further undercut by George’s pronouncement that officers need not “delay their fire until a suspect turns his weapon on them” when a person “reasonably suspected of being armed” makes “a furtive movement,” a “harrowing gesture,” or even a “serious verbal threat.” Id. at 838. This passage stands for the proposition that the use of deadly force can be justified by an action less threatening than pointing a gun directly at an officer. Combining this principle with the case’s holding that deadly force is not reasonable if an armed individual does not manipulate his gun, the use of deadly force against a person armed with a gun (or reasonably suspected of being so armed) becomes reasonable somewhere along the spectrum of actions between not manipulating the gun and pointing the gun at an officer. One would search in vain, however, to find the point at which that occurs in George. There simply was no reason to reach that issue based on the factual assumptions the court was required to make. Accordingly, George could not have put Deputy Gelhaus on notice that Andy’s actions did not cross the threshold—wherever it may lie—at which the use of deadly force becomes reasonable. He may have been mistaken in his assessment, but he would not have known it from reading George.
Once this is understood, the additional aggravating factors of George become immaterial. Nevertheless, it bears mentioning that the majority greatly understates the potential danger Andy posed as perceived by Deputy Gelhaus. As explained, Deputy Gelhaus reasonably believed that Andy was carrying an AK-47. With narrow exceptions, possession of such a weapon is a crime in California. Cal. Penal Code § 30605(a). Considering the undisputed destructive capabilities of an AK-47, the prevalence of weapons-related violent crimes in the area, and the fact that local gang members were known to use weapons against police to gain respect, the suspected crime cannot be considered mild. Indeed, in enacting this prohibition, the California legislature declared that “the proliferation and use of assault weapons poses a threat to the health, safety, and security of all citizens of [California].” Id. § 30505(a).1 By contrast, the possible crime in George was less threatening. As we observed, the victim’s wife, who had made the 911 call, “was unscathed and not in jeopardy when deputies arrived.” George, 736 F.3d at 839. Furthermore, her husband “was not in the vicinity,” and was instead “said to be on the couple’s rear patio.” Id.
Nor can the majority rely on George because it established a rule that the use of deadly force without an objective threat is unreasonable, because this framing commits the sin for which the Supreme Court repeatedly has admonished the lower federal courts: it “defme[s] clearly established law at a high level of generality.” Ashcroft v. al-Kidd, 563 U.S. 731, 742, 131 S.Ct. 2074, 179 L.Ed.2d 1149 (2011); see also Mullenix v. Luna, — U.S.—, 136 S.Ct. 305, 309, 193 L.Ed.2d 255 (2015) (“The general principle that deadly force requires a sufficient threat hardly settles this matter”). The operative inquiry instead is whether there is a case that would have given notice to Deputy Gelhaus at the time of the incident that the circumstances he faced were not sufficiently threatening to warrant the use of deadly force. See White, 137 S.Ct. at 552 (reversing denial of qualified immunity because the court “failed to identify a case where an officer acting under similar circumstances as [the defendant] was held to have violated the Fourth Amendment” (emphasis added)). For the reasons already explained, George is not such a case.2
B.
The second case cited by the majority, Harris v. Roderick, 126 F.3d 1189 (9th Cir. 1997), also fails to live up to its announced billing. Harris arose from the Ruby Ridge siege and involved the use of deadly force by a Federal Bureau of Investigation (FBI) sniper against the plaintiff, Harris. 126 F.3d at 1193-94. Following a shootout between United States Marshals and a group of armed civilians including Harris, “the FBI dispatched a special unit designed to deal with crisis situations,” which included snipers. Id. at 1193. For purposes of this encounter alone, the FBI, in collaboration with the Marshal Service, rewrote its Standard Rules of Engagement. Id. The new rules displaced the requirement that, deadly force be used only when the target “presents an immediate risk of death or great bodily harm to the agent or another person” in favor of an instruction that “any armed adult male” “in the vicinity of the Weaver cabin could and should be killed.” Id. (emphasis omitted).
According to the complaint, the events leading up to the shooting of Harris unfolded as follows. The day after the initial shootout, Harris accompanied Randy Weaver, the owner of the cabin under siege and the person upon whom the Marshals were attempting to serve an arrest warrant when the shootout erupted, to a shed on the property “to help minister to the body of Weaver’s dead son,” who had been killed in the shootout. Id. at 1193, 1203. While Weaver was opening the shed, an FBI sniper stationed “on a hill overlooking the Weaver cabin” shot Weaver in the back. Id. at 1193. Harris was armed at this point, but “made no aggressive move of any kind.” Id. at 1203. The group immediately ran back to the cabin, where Weaver’s wife, Vickie, was holding the door open. Id. at 1193. As Harris was entering the cabin, the sniper “fired a second shot in an effort to kill both Harris and Vickie.” Id. at 1193-94. “The bullet passed through the clear glass in the open door, striking Vickie in the head, and after passing through her, hit Harris in the upper arm and chest.” Id. at 1194.
The facts of our case are far afield from those in Harris. Unlike Deputy Gelhaus, the FBI sniper was “perched safely on a hill” when he started shooting. Unlike Andy, Harris was not turning to face the agent but rather was fleeing back into the cabin at the time he was shot. Finally, although Harris was armed, there was no indication that his weapon made any movement in the sniper’s direction before the latter resorted to deadly force. Indeed, the facts as alleged made it clear that the sniper shot Harris solely because he was armed, and that was the rule that the case established: “Law enforcement officials may not kill suspects who do not pose an immediate threat to their safety or to the safety of others simply because they are armed.” Id. at 1204 (emphasis added).
We, of course, are not dealing with a situation in which Deputy Gelhaus shot Andy merely because he was armed. Knowing that he could not use deadly force just because Andy was holding a gun would not tell Deputy Gelhaus what the Constitution required when Andy, instead of following the command to drop the gun, turned to face Deputy Gelhaus and the barrel of the rifle began to rise. Harris did not address such a circumstance, or even a similar circumstance, and so could not have given Deputy Gelhaus notice one way or the other as to the reasonableness of his actions. It therefore is inapposite to the question we face hi this case.
C.
The majority’s final case, Cumow ex rel. Cumow v. Ridgecrest Police,- 952 F.2d 321 (9th Cir. 1991), is even less helpful in this analysis because, as the majority admits, there was evidence in that case that the victim was unarmed at the time police began shooting at him. Id. at 323. But that is not all: a witness to the shooting stated that the victim not only was not armed, but had not even reached for a nearby gun when an officer shot him in the back. Id. In the most favorable light, the victim was merely sitting in his home, with his back to the officer and a gun in the vicinity. A rule that deadly force is unreasonable in those circumstances says nothing about the propriety of such force when the person is armed and facing the officer and the gun is beginning to rise. The majority suggests that this case provided “ ‘fair notice’ that the use of deadly force is unreasonable where the victim does not directly threaten the officer with the gun.” Not only is this interpretation inconsistent with George’s admonition that officers are not always required “to delay their fire until a suspect turns his weapon on them,” 736 F.3d at 838, it is also inappropriate because the undisputed facts here do establish a direct threat to the officer. Thus, Curnow is off-point as well.
III.
The disputed facts the majority points to—whether Andy looked backwards at the officers, whether Deputy Gelhaus yelled at Andy to drop the gun more than once, whether the patrol car chirped more than once, whether Andy held the gun in his right or left hand, and the angle between the ground and Deputy Gelhaus at which Andy pointed his gun—are simply not material to the qualified immunity analysis. Taking together the district court’s findings and undisputed facts, this case involves the use of deadly force against a hooded individual armed with a replica assault rifle indistinguishable from a real one, who turned to face' an officer while raising the rifle after the officer had activated his patrol car lights and siren and yelled at the individual to drop the rifle. These facts are not sufficiently similar to the facts of George, Harris, or Cur-now to have put Deputy Gelhaus on notice that his use of deadly force violated Andy’s Fourth Amendment right to be free from excessive force. See White, 137 S.Ct. at 552. Without these cases, the majority is left only with the statement it cites at the beginning of its clearly established law analysis: that we may deny qualified immunity “in novel circumstances.” Hughes v. Kisela, 862 F.3d 775, 783 (9th Cir. 2016). It is doubtful how much of this statement,- if any, has survived the Supreme Court’s intervening decision in White. See 137 S.Ct. at 552 (stating that the Tenth Circuit’s observation that the case “present[ed] a unique set of facts and circumstances.... should have been an important indication ... that [the officer’s] conduct did not violate a clearly established right” (internal citation and quotation marks omitted)). To the extent it retains any vitality, it likely would be confined to those cases where the officer’s conduct is an “obvious” violation of a constitutional right. Id., quoting Brosseau v. Haugen, 543 U.S. 194, 199, 125 S.Ct. 596, 160 L.Ed.2d 583 (2004) (per curiam); see also Deorle v. Rutherford, 272 F.3d 1272, 1286 (9th Cir. 2001) (“When ‘the defendant[’s] conduct is so patently violative of the constitutional right that reasonable officials would know without guidance from the courts’ that the action was unconstitutional, closely analogous pre-existing case law is not required to show that the law is clearly established’ ” (quoting Mendoza v. Block, 27 F.3d 1357, 1361 (9th Cir. 1994))).
This assuredly is not such “an obvious case.” Brosseau, 543 U.S. at 199, 125 S.Ct. 596. As shown by the majority’s painstaking evaluation of the objective reasonableness of Deputy Gelhaus’s use of force, this case is not obvious, but clearly quite close. Whether Deputy Gelhaus acted unreasonably turns on such minute details as how high the gun barrel had risen, whether it might have been feasible to give a warning, and just how aggressive Andy’s turning motion was. By contrast, cases found to be “obvious” involve much clearer constitutional transgressions. See, e.g., Hope v. Pelzer, 536 U.S. 730, 734-35, 122 S.Ct. 2508, 153 L.Ed.2d 666 (2002) (reversing grant of qualified immunity where a prisoner was handcuffed to a “hitching post” without a shirt for seven hours “while the sun burned his skin,” during which time “he was given water only once or twice and was given no bathroom breaks” and a guard “taunted [him] about his thirst” by giving water to some dogs, bringing the water cooler near the prisoner, and then intentionally spilling all the water on the ground). Our case is not the “rare” one “in which the constitutional right at issue is defined by a standard that is so ‘obvious’ that we must conclude ... that qualified immunity is inapplicable, even without a case directly on point.” A.D. v. Cal. Highway Patrol, 712 F.3d 446, 455 (9th Cir. 2013), quoting Hope, 536 U.S. at 740-41, 122 S.Ct. 2508. Accordingly, the district court’s denial of immunity cannot be affirmed on this basis either.
IV.
Deputy Gelhaus misjudged the threat that Andy posed, and Andy’s death is the heartbreaking result of that miscalculation. In circumstances like these, it is imperative that we do justice. But justice does not invariably require punishing the officer. A reasonable mistake of law or fact is not enough to impose liability. Pearson, 555 U.S. at 231, 129 S.Ct. 808. The law affords relief only when an officer transgresses a boundary clearly established by precedent at the time he acts. If no such case exists, the officer cannot be held liable even if his conduct, the court believes in retrospect, may be unreasonable.
This is the situation that we face. The facts of the cases that the majority relies on to reach the opposite conclusion are materially different from the real facts before us. Those cases therefore could not have given Deputy Gelhaus notice that using deadly force against Andy would violate his constitutional right. Although all are sympathetic to Andy’s family, as anyone should be, I am duty-bound to conclude that we must provide Deputy Gel-haus with the “breathing room to make reasonable but mistaken judgments about open legal questions” that qualified immunity affords him. al-Kidd, 563 U.S. at 743, 131 S.Ct. 2074. For these reasons, I dissent.
. That Deputy Gelhaus might have reasonably suspected that Andy was committing a nontrivial crime also bears on the first factor in the Graham excessive force analysis, contrary to the majority’s assertion that this factor "weights] clearly in Andy's favor.” Because I conclude that Deputy Gelhaus is entitled to immunity because it was not clearly established that his conduct was unconstitutional, however, I would not speculate on whether a reasonable jury could find his use of deadly force to be objectively unreasonable, and do not do so here. See Pearson v. Callahan, 555 U.S. 223, 236, 129 S.Ct. 808, 172 L.Ed.2d 565 (2009).
. The majority objects to this paragraph as employing a “fictitious frame” of its argument. There is no cause for alarm, however. I am simply pointing out that reading George to establish a more general rule is no more helpful to the majority's analysis than the actual facts of the case.