concurring in part and dissenting in part.
Thane Carl Chew was stopped by a Los Angeles Police Department (“LAPD”) officer for a traffic violation and identified himself by presenting his driver’s license. When the officer returned to his car to check Chew’s record, Chew ran away. The officer pursued Chew, who scaled several fences during the chase before ultimately hiding in a scrapyard. During the subsequent search, which involved a number of officers and several K-9 units, Officer Bunch unleashed a police dog named Volker to find Chew. Defendants stipulated for summary judgment purposes that Chew “tried to surrender peacefully once he realized he had been found” by Volker. Defendants’ Reply to Plaintiffs Opposition to Defendants’ Motion for Summary Judgment at 2 [hereinafter “Defendants’ Reply”]. Nevertheless, Volker seized Chew by biting him hard and holding on to him. Volker was out of Officer Bunch’s sight when he located Chew and initiated his attack. Volker’s attack left Chew with severe bite injuries on his arm and torso. At the time Officer Bunch released Volker, Bunch knew only that there were three outstanding warrants for Chew’s arrest on unspecified felony charges. Officer Bunch had no reason to believe Chew was armed.
The City of Los Angeles and the four individual officers charged with the policy-making responsibility for the LAPD canine policy (then-Chief Daryl Gates, Captain Patrick McKinley, and Sergeants Donn Yarnall and Mark Mooring),1 moved for summary *1452judgment on the ground that the LAPD canine policy was constitutional. In their summary judgment papers, they made no reference to the facts of Chew’s seizure by Volker. They merely submitted the LAPD’s written canine policy and asked the court to declare it constitutional as a matter of law.2 The individual policymaking defendants also argued that they were entitled to summary judgment on the ground of qualified immunity. The district court awarded these defendants summary judgment, which Chew appealed after the jury rendered a verdict in his favor against Officer Bunch.
I
The Constitutionality of the LAPD Canine Policy
A
This appeal requires us to judge the constitutionality of the LAPD canine policy under a well established principle of excessive force law: Deadly force may not be used to prevent the escape of a suspect unless “the officer has probable cause to believe that the suspect poses a significant threat of death or serious physical injury to the officer or others.” Tennessee v. Garner, 471 U.S. 1, 3, 105 S.Ct. 1694, 1697, 85 L.Ed.2d 1 (1985).
In Gamer, police shot and killed a burglary suspect trying to escape by climbing a backyard fence. The suspect, Edward Garner, an eighth grader, was 15. He was unarmed. Ten dollars and a stolen purse were found on his body. See Garner, 471 U.S. at 3-4, 105 S.Ct. at 1697-98. The issue presented in Gamer was the constitutionality of a Tennessee statute that permitted the use of “all necessary means” to apprehend fleeing suspects, whether dangerous or not. Id. at 4, 105 S.Ct. at 1697-98. The Supreme Court held that the Tennessee statute violated the Fourth Amendment insofar as it permitted the use of deadly force to prevent the escape of a felony suspect who “poses no immediate threat to the officer and no threat to others.” Id. at 11, 105 S.Ct. at 1701. In explaining its decision, the Court said: “[t]he use of deadly force to prevent the escape of all felony suspects, whatever the circumstances, is constitutionally unreasonable. It is not better that all felony suspects die than that they escape.” Id. (emphasis added).
In this appeal, we must apply Gamer to decide the constitutionality of a municipal policy which permits the use of police dogs to prevent the escape of all felony suspects, regardless whether the officer “has probable cause to believe that the suspect poses a significant threat of death or serious physical injury to the officer or others.” Id. at 3, 105 S.Ct. at 1697. See Los Angeles Police Department Canine Unit Manual at 5, 9, 38, 39, 41, 44, 62 [hereinafter “LAPD Canine Manual”].3
The LAPD canine policy tracks the Tennessee statute in that it does not differentiate between dangerous and non-dangerous felony suspects. The only material difference between the LAPD policy and the Tennessee statute is that the statute authorized the use of “all means necessary,” whereas the policy authorizes only the use of dogs. If the use of the LAPD dogs constitutes the use of deadly force, Gamer would stand in the way of defendants’ motion for summary judgment because the LAPD canine policy would be unconstitutional, as the Tennessee statute was unconstitutional, for permitting the use of deadly force to prevent the escape of felony suspects who do not pose “a signifi*1453cant threat of death or serious physical injury to the officer or others.” Garner, 471 U.S. at 3, 105 S.Ct. at 1697. Accordingly, defendants’ summary judgment motion in its present form must be denied because the critical question whether the use of LAPD dogs, as trained and deployed, constitutes the use of deadly force cannot be decided as a matter of law on the summary judgment record before us.
Deadly force is force that creates a substantial risk of causing death or serious bodily harm.4 A gun plainly qualifies as deadly force. Whether a dog also qualifies depends on how it is trained to behave when eonfront-ing a suspect. For example, a dog trained to find a suspect and bark until the dog’s handler arrives would plainly not qualify as an instrument of deadly force.5 But a German Shepherd that is behaviorally conditioned to go directly for a suspect’s jugular would surely qualify as an instrument of deadly force.
Whether a particular instrument of force qualifies as an instrument of deadly force is a question of fact.6 Indeed, whether particular dogs as trained and deployed qualify as deadly weapons has been uniformly treated as a question of fact for the jury.7 On the eviden-*1454tiary record before us, the question whether or not the LAPD dogs, as trained to bite and hold suspects, constitute instruments of deadly force is an issue of fact that cannot be decided at summary judgment.
In their summary judgment papers, defendants fail to address this question: whether the dogs they train to bite and hold suspects are deadly force. They proffer no evidence whatsoever on the magnitude of the risk that their dogs will kill or seriously injure suspects. They provide no empirical data, for example, showing how frequently LAPD dogs bite the suspects they confront or the seriousness of the injuries suspects sustain when they are bitten. In short, defendants ask us to declare their canine policy constitutional as written without the benefit of evidence on the risk of harm the dogs represent.
In opposing the summary judgment motion, Chew proffered considerable evidence that the LAPD dogs, as trained and deployed, create a substantial risk of serious injury to suspects. He cites deposition testimony of Sergeant Yarnall that Volker was not trained to grip only a suspect’s arm or leg, but instead was trained to bite whatever part of the suspect’s body was necessary for an effective hold. See Chew Supplemental Memorandum of Points and Authorities in Opposition to Defendants’ Motion for Summary Judgment at 14 (citing Deposition of Yarnall at 33-35). Chew also cites deposition testimony of then-Chief Daryl Gates that Gates was “[v]ery much” aware that Volker, as trained, could maim and even kill a suspect, Deposition of Daryl Gates at 21,8 as well as testimony of Officer Bunch acknowledging that dog bites can be “fatal” or cause “serious and permanent” injuries, Deposition of Bunch at 30-32. Furthermore, Bunch acknowledged that Chew’s injuries were “pretty severe,” with “serious lacerations.” Id. at 126. When read in a light most favorable to Chew, this evidence raises a genuine issue of material fact as to whether the LAPD dogs, as trained and deployed pursuant to the LAPD canine policy, create “a substantial risk of causing death or serious bodily harm.” Model Penal Code § 3.11(2).9
In sum, the Supreme Court held in Gamer that by authorizing the use of deadly force against all fleeing felons, the Tennessee statute was based on an anachronistic brightline distinction between felons and misdemean-ants. As the Court made clear, the critical factor in determining when the use of deadly force is warranted is not whether the law classifies the suspect’s crime as a felony or as a misdemeanor, but whether the suspect poses a significant threat of harm to others. See Garner, 471 U.S. at 11, 14, 19-20, 105 S.Ct. at 1701, 1702-03, 1705-06. The LAPD canine policy relies upon this same outdated and constitutionally irrelevant distinction between felonies and misdemeanors. See LAPD Canine Manual at 5, 9, 38, 62. Thus, if the LAPD dogs, as trained and deployed, constitute instruments of deadly force, then the LAPD canine policy violates the Fourth Amendment in exactly the same manner as did the Tennessee statute because it fails to limit the application of deadly force to those suspects who pose a significant threat to *1455others. Whether the LAPD dogs constitute instruments of deadly force is a genuine issue of fact which precludes summary judgment on the theory advanced by defendants — that the LAPD canine policy, as written, is constitutional as a matter of law.
B
It is important to make clear that even if the unleashing of Volker constituted the application of deadly force against Chew, it would not have been excessive force if Officer Bunch had probable cause to believe that Chew posed a significant threat of serious bodily harm to themselves or to others. If he had probable cause to believe Chew posed such a threat, the use of deadly force to prevent his escape would have been objectively reasonable under Gamer.10 In that scenario, Chew would have no standing to challenge the constitutionality of the LAPD canine policy because he would have suffered no constitutional injury.
Whether Officer Bunch had probable cause to believe Chew posed a significant threat of serious physical harm to others is an issue of fact that cannot be resolved on summary judgment.11 Officer Bunch knew that Chew was a felony suspect, but the nature of the felonies was unknown. And after Gamer, Chew’s suspected felony status alone was insufficient to create the probable cause necessary to use deadly force. See Garner, 471 U.S. at 14, 105 S.Ct. at 1703 (“[T]he assumption that a ‘felon’ is more dangerous than a misdemeanant [is] untenable.”). The evidence in the summary judgment record does not establish as a matter of law that Chew posed any greater threat than Edward Garner did. In neither ease did the officer have reason to believe the suspect was armed. In both cases, there was probable cause to believe the fleeing suspect was a felon, but no reason to believe he “pose[d] a significant threat of death or serious physical injury to the officer or others.” Id. at 1452. Finally, in neither case did the circumstances of the attempted escape add anything to suggest that the suspect was dangerous. Garner ran from the home where the burglary was committed through the backyard to a fence, which he had begun to climb when he was shot. Chew made a break for it while talking peacefully to the officer who had stopped him for a traffic violation, jumped several fences, and was hiding in a scrapyard when he was found by Volker. Despite Chew’s attempt to surrender peacefully, see Defendants’ Reply at 2, Volker seized Chew by biting him repeatedly. On these facts, we cannot grant summary judgment to defendants on the ground that Chew posed the requisite threat as a matter of law any more than we could have decided that Garner posed the same threat as a matter of law.12
C
In echoing defendants’ refrain that the LAPD’s canine policy is constitutional as a *1456matter of law, Judge Trott raises the specter that our decision today may lead to “the elimination of the use of police dogs in this circuit to find hiding felony suspects,” no matter how the dogs are trained or deployed. Opinion of Judge Trott at 1463. To dramatize the importance of well-trained police dogs to law enforcement and the public safety, Judge Trott recounts the story of Pascha, the police dog that helped the LAPD apprehend a career violent criminal known as the “Balcony Burglar.” Id. at 1473. However, the heroic story of Pascha has no bearing on the question before us: whether on this record defendants are entitled to summary judgment.
Like Judge Trott, I recognize that Pascha was a great hero. Also like Judge Trott, “I do not believe that the deployment per se of police dogs to find hiding felony suspects is unreasonable.” Id. at 1472.13 However, I don’t buy Judge Trott’s inflammatory rhetoric and doomsday prediction that denying defendants summary judgment will prevent the use of all police dogs in all instances. Lest my opinion be misinterpreted, I emphasize two points.
First, I am not saying that all police dogs, regardless of how they are trained and deployed, constitute deadly force. I am saying only that whether the police dogs, trained by the LAPD to bite and hold, are instruments of deadly force is an issue of fact that cannot be resolved on the summary judgment record before us. There are obviously many ways to train and deploy a police dog so that it is not deadly force, one of which is the “find and bark” policy that is apparently once again in use by the LAPD. See supra note 5. Whether or not a dog is permitted to roam out of sight of its handler, as Volker was, would be another factor to consider.
Second, I am not saying that the use of a police dog, even if trained to deliver deadly force, is necessarily unreasonable. I am saying only that under Gamer, if a dog is trained to deliver deadly force, then the Fourth Amendment requires that its use be limited to those situations where there is probable cause to believe that the suspect poses a significant threat of death or serious bodily harm. In the case of the Balcony Burglar, which involved the pursuit and apprehension of a person suspected of committing numerous murders, rapes, and robberies, Pascha’s use, whether deadly force or not, plainly did not violate the Fourth Amendment.
II
The City’s Monell Liability
In resting its summary judgment motion solely on the argument that the LAPD’s “find, bite, and hold” canine policy is constitutional as a matter of law, the City conceded that Officer Bunch acted in accordance with LAPD policy when he unleashed Volker to hunt down and seize Chew. See Opinion of Judge Reinhardt at 1444 n. 13. On appeal, following the jury verdict that Officer Bunch violated Chew’s Fourth Amendment rights, the City has changed its tune. It now argues that Officer Bunch’s actions constituted nothing “more than ... a random act or isolated event that resulted in a constitutional rights violation.” Appellees’ Br. at 31. In other words, it argues that it is not hable for Chew’s dog bite injuries because even if Volker’s use in this case constituted an unreasonable seizure in violation of the Fourth Amendment, the City was not the “moving force” behind the violation. Id. at 32 (stating that “[tjhere is no evidence of any policy to engage in the actions attributed to the officer”). See Monell v. Department of Social Servs. of City of New York, 436 U.S. 658, 694, 98 S.Ct. 2018, 2037, 56 L.Ed.2d 611 (1978) (a city is liable only if city policy “causes” or is “the moving force” behind a constitutional violation).
The City’s appellate position is a 180-de-gree reversal from its stipulated position below that Officer Bunch’s use of Volker to seize Chew complied fully with the LAPD canine policy. I agree with Judge Reinhardt that we must hold the City to the concession it made for summary judgment purposes, see *1457Opinion of Judge Reinhardt at 1444-45,14 and reject the City’s new Monell position — which it takes for the first time on appeal — distancing itself from Officer Bunch’s conduct.15
Ill
Qualified Immunity16
The individual defendants (Gates, McKinley, Yarnall, and Mooring), who had the poli-cymaking responsibility for the LAPD canine policy, argue that even if the policy did violate the Fourth Amendment, they would have no liability for Chew’s injuries because they enjoy qualified immunity. The doctrine of qualified immunity shields “government officials performing discretionary functions ... from liability for civil damages insofar as their conduct does not violate clearly established statutory or constitutional rights of which a reasonable person would have known.” Harlow v. Fitzgerald, 457 U.S. 800, 818, 102 S.Ct. 2727, 2738, 73 L.Ed.2d 396 (1982).
A
The doctrine of qualified immunity requires a two-prong inquiry: “1) Was the law governing the official’s conduct clearly established? 2) Under that law, could a reasonable officer have believed the conduct was lawful?” Act Up!/Portland v. Bagley, 988 F.2d 868, 871 (9th Cir.1993). The first prong involves a pure question of law for the court to decide, Mendoza v. Block, 27 F.3d 1357, 1360-61 (9th Cir.1994); the second prong involves a mixed question of fact and law requiring the application of the clearly established law to the facts of the particular case.
1
The first prong of the qualified immunity test requires us to ask whether the Fourth Amendment principles governing the use of police dogs to prevent suspects from escaping were clearly established at the relevant time. In Mendoza, also a dog bite case, the officers argued that the law governing the use of police dogs was not clearly established because of the paucity of case law specifically addressing the use of dogs in police work. We rejected that argument:
The deputies argue that the law guiding the use of police dogs at the time Mendoza was bitten was not clearly established. They point to the lack of case law addressing the use of dogs to locate fleeing suspects, and argue that under the circumstances they could not reasonably have known how to pattern their conduct so as to avoid liability.
* * * * * *
This does not mean, however, that there was no clearly established law that would indicate to the deputies that a constitutional right might be violated when using a dog during an arrest. Mendoza argues that the use of a police dog in this case constituted excessive force. Officers have a considerable amount of guidance from the courts in the permissible use of force while making an arrest. It is clearly established that “[t]he use of excessive force by police officers in an arrest violates the arrestee’s Fourth Amendment right to be free from an unreasonable seizure.” White v. Pierce County, 797 F.2d 812, 816 (9th Cir.1986). “The reasonableness of *1458force is analyzed in light of such factors as the requirements for the officer’s safety, the motivation for the arrest, and the extent of the injury inflicted.” Id. This analysis applies to any arrest situation where force is used, whether it involves physical restraint, use of a baton, use of a gun, or use of a dog. We do not believe that a more particularized expression of the law is necessary for law enforcement officials using police dogs to understand that under some circumstances the use of such a “weapon” might become unlawful. For example, no particularized case law is necessary for a deputy to know that excessive force has been used when a deputy sics a canine on a handcuffed arrestee who has fully surrendered and is completely under control. An officer is not entitled to qualified immunity on the grounds that the law is not clearly established every time a novel method is used to inflict injury. See Anderson, 483 U.S. at 640 [107 S.Ct. at 3039].
We therefore hold that the deputies’ use of the police dog is subject to excessive force analysis, and that this law is clearly established for purposes of determining whether the officers have qualified immunity.
Id., 27 F.3d at 1362 (emphasis added).
Mendoza is controlling authority that the first prong of the qualified immunity test is satisfied in this case: the use of LAPD dogs is governed by established principles of excessive force law. It is a basic tenet of excessive force law that deadly force may be used only if necessary to prevent the escape of a suspect who poses “a significant threat of death or serious physical injury.” Garner, 471 U.S. at 3, 105 S.Ct. at 1697. Thus, Gamer held unequivocally that deadly force cannot be used to prevent the escape of all felons regardless of the danger they pose to others. Deadly force may only be used to prevent the escape of some felons — those who pose a serious threat to the safety of others.
2
The second prong of the qualified immunity inquiry requires us to apply the clearly established principles of excessive force law to the facts of the particular use of force in this case. Here we must ask whether, in light of the clearly established law governing the use of force, a reasonable officer in defendants’ policymaking position could have believed it constitutional to adopt a policy permitting the use of dogs to bite and hold felony suspects, regardless whether they posed “a significant threat of death or serious physical injury to the officer or others.” Id. at 1452.
In Gamer, the Court held that a Tennessee statute violated the Fourth Amendment to the extent it permitted the use of deadly force to prevent the escape of felons who did not pose a significant threat of serious harm to others. After Gamer, it could not have been clearer to the defendants that the LAPD canine policy was unconstitutional to the extent it authorized the use of deadly force to prevent the escape of felony suspects who did not pose the requisite threat. Because the LAPD policy does not limit the use of police dogs to felony suspects reasonably believed to be dangerous, the question whether a reasonable officer could have believed the LAPD canine policy was constitutional turns on whether a reasonable officer could have believed the use of LAPD dogs such as Volker did not constitute the use of deadly force.
That is a question we cannot decide on the summary judgment record before us. As I have already explained, a genuine issue of fact exists whether the LAPD dogs, as trained and deployed, constitute deadly force. See supra at 1454-55. For emphasis, I will repeat that defendants proffered no evidence in support of their summary judgment motion regarding the danger posed by the LAPD dogs, such as evidence of the type, frequency, and severity of injuries they have inflicted. On the other hand, Chew offered evidence such as the deposition of ex-Chief Gates, who testified that he was “[v]ery much” aware that Volker, as trained, could maim and even kill a suspect. See supra at 1454. Accordingly, because we cannot decide whether a reasonable policymaker could have believed that the LAPD dogs were not deadly force, we cannot decide that defendants are entitled to summary judgment on the *1459ground they rely upon in their motion — that a reasonable officer could have believed the LAPD canine policy was constitutional as written.17
B
Defendants’ claim of qualified immunity is based almost exclusively on the Sixth Circuit’s decision in Robinette v. Barnes, 854 F.2d 909 (6th Cir.1988). In that case, a police dog bit a suspect’s neck and killed him, an event that the Sixth Circuit said was an “extreme aberration.” Id. at 912.18 On the record before it, the court held that that particular dog did not constitute deadly force, and that the use of non-deadly force was reasonable under the totality of circumstances of the case. See id. at 912-13. Alternatively, the court held that even if the dog was a deadly weapon, its use would have been reasonable because the suspect “threatened [the officer’s] safety and the safety of the other officers present.” Id. at 913. Defendants here argue they are entitled to qualified immunity because Robinette was the only relevant federal court of appeals opinion on the books when they adopted the LAPD canine policy.
In Mendoza, we discounted Robinette as authority on the issue of qualified immunity in other dog bite cases because it was “fact-specific.” See Mendoza, at 1361-62 (“It is therefore questionable whether fact-specific case law guided the deputies’ use of a police dog when they arrested Mendoza.”). We did not regard Robinette as an automatic guarantee of qualified immunity to officers in other dog bite cases. Instead, we affirmed the district court’s award of qualified immunity because the district court had made a finding of fact, following an evidentiary hearing, that “Mendoza was fleeing arrest for a bank robbery ... [t]he deputies believed he was armed ... [and] the deputies could reasonably have believed he posed a danger not only to themselves but also to the property owners.” Id. at 1362. Under these findings of fact, the use of police dogs to prevent Mendoza’s escape would satisfy the requirements of Gamer even if the dogs constituted deadly force. See Garner, 471 U.S. at 11-12, 105 S.Ct. at 1701-02.
My colleagues seem to think that after Robinette, a reasonable officer could have believed it was constitutional to deploy German Shepherds trained to bite and hold even non-dangerous suspects. Implicit in their reasoning is the assumption that it follows, as the night the day, that once a particular dog is judicially determined to be an instrument of non-deadly force, a reasonable officer could believe that no dog is an instrument of deadly force, however trained and deployed.
My analysis rejecting defendants’ claim of qualified immunity based on Robinette is supported not only by Mendoza, but also by Marley v. City of Allentown, 774 F.Supp. 343 (E.D.Pa.1991), aff'd without opinion, 961 F.2d 1567 (3d Cir.1992). Because the Marley court applied qualified immunity analysis to the precise issue presented in this case — a claim of qualified immunity based on Robinette—it is instructive to lay out Marley in some detail.
Exactly as in this case, the officer in Marley staked his claim of qualified immunity on Robinette, moving for a directed verdict on the ground that the Sixth Circuit “said that the use of a police dog was not deadly force.” Transcript of district court proceedings in Marley v. City of Allentown, July 2, 1991, at 188-89. The district court denied the motion, and the case went to the jury with the instruction that “[d]eadly force means force that’s likely to kill or likely to seriously injure. You heard about the training of the *1460police dog and how it’s handled. You need to determine whether or not that was deadly force under that definition.” Id. at 200. Following the jury’s verdict against the officer, he moved for judgment notwithstanding the verdict, again arguing that he was entitled to qualified immunity on the basis of Robinette.
In denying the motion, the Marley court used the same level of generality we later approved in Mendoza to determine whether the applicable law was clearly established at the time of the officer’s conduct. The Marley court refused to limit itself to police dog cases and instead considered the qualified immunity claim in light of the generic excessive force principles laid down in eases such as Gamer. The Marley court observed that “[i]n Robinette, the ... Sixth Circuit merely applied [Gamer] to the facts of the case before it” and concluded that there was no Fourth Amendment violation. 774 F.Supp. at 345. The Marley court then explained that the application of Gamer to the facts of its dog bite case led to a different result than the result in Robinette. Accepting the jury finding that the dog represented deadly force, the court held that in light of Gamer, no reasonable officer could have believed it was lawful to use the dog to seize a fleeing suspect who “one could reasonably conclude ... posed no threat to the officer.”. Id.
Marley’s qualified immunity methodology — looking beyond dog bite cases to generic principles of excessive force jurisprudence and then applying those principles to the facts of the particular case — is not only correct, it is mandated by our recent decision in Mendoza. Mendoza teaches that under the first prong of qualified immunity analysis, generic principles of excessive force law apply regardless whether the particular instrumentality of force involves “physical restraint, use of a baton, use of a gun, or use of a dog.” Mendoza, 27 F.3d at 1361-62. Garner provided explicit guidance for the LAPD policymakers that if they train dogs to bite and seize suspects in a way that creates a “substantial risk of causing death or serious bodily harm,” Model Penal Code § 3.11(2), they had better restrict their use to suspects reasonably believed to be dangerous.
Throughout Part ÍV.B.2 of Judge Reinhardt’s opinion, my colleagues scramble their qualified immunity analysis by failing to consider separately the two independent prongs of the qualified immunity test. See supra at 1457, 1458. They do so because they fail to recognize that prong one — the “clearly established” law prong — raises a pure question of law that does not turn on the particular facts of the case; the facts do not enter into the analysis until we reach the second prong, where we apply the established law — in this case generic principles of excessive force law — to the facts. This analytical flaw is demonstrated by the way they criticize my use of Mendoza and Marley as precedent. See Opinion of Judge Reinhardt at 1448-49. I cite both cases as authority that the law governing all excessive force cases, regardless of the instrument used to apply the force, is “clearly established.” Indeed, I cite Mendoza as controlling circuit authority that the first prong of the qualified immunity inquiry is satisfied in excessive force cases by the generic principles established in Gamer. Mendoza should foreclose any further discussion on the first prong. In criticizing my limited use of Mendoza and Marley as precedent, my colleagues discuss the facts of those cases, as well as the facts of Chew’s case, in a way that injects facts into what should be a pure question of law: “Was the law governing the official’s conduct clearly established?” Act Up!, 988 F.2d at 871. The result is a fundamental and far-reaching error in deciding the qualified immunity issue at the wrong level of generality.19
*1461My colleagues also fail to recognize that I make use of Mendoza and Marley as authority that the second prong of the qualified immunity test is fact-specific. In Mendoza, the district court granted qualified immunity on the basis of facts it found at a pre-trial evidentiary hearing; in Marley, the jury returned a verdict against the officer. The district court subsequently denied the officer’s motion for JNOV. In other words, even though both were dog bite cases, the Mendoza and Marley courts reached different results on the qualified immunity issue by applying established generic principles of excessive force law to the particular facts of the case.20
C
Another reason offered by my colleagues for cloaking the policymaking officers with qualified immunity is that the LAPD canine policy was “a longstanding official policy, which was well-known and similar to the policies employed in many police departments throughout the nation, none of which had been judicially questioned.” Opinion of Judge Reinhardt at 6978. First, there is no basis in the record for the majority’s assertion that the policy was “well-known” and “similar” to other canine policies. Second, an unconstitutional police practice does not metamorphose into a constitutional practice by escaping judicial scrutiny over time. The happenstance that a particular police practice has not yet been “judicially questioned” is meaningless. Since courts do not sua sponte review police practices, the fact that a practice has not yet been “judicially questioned” may simply mean that an actual case or controversy that squarely presents the issue has not found its way into a published opinion.
To give the majority its due, it at least acknowledges that if a police policy is not “ ‘meaningfully distinguishable’ ” from a policy that has been declared unlawful, then qualified immunity may not be invoked no matter how widespread and longstanding the policy is. Opinion of Judge Reinhardt at 1449-50 (quoting Wood v. Ostrander, 879 F.2d 583, 592 (9th Cir.1989)). What the majority fails to explain is how the use of LAPD dogs — trained to roam beyond the handler’s control, bite hard, and seize even non-resisting, non-dangerous suspects — is “meaningfully distinguishable” for Fourth Amendment purposes from the use of guns or other deadly weapons against non-dangerous suspects.
From the victim’s perspective, deadly force is deadly force, regardless of whether the force comes from the muzzle of a gun or the muzzle of a dog. Nor does it matter from the perspective of the law. The Fourth Amendment requires that an officer’s use of force be reasonable regardless whether the force involves “physical restraint, use of a baton, use of a gun, or use of a dog”. Mendoza, 27 F.3d at 1362. Garner specifically limits the use of deadly force regardless of the instrumentality used. Garner, 471 U.S. at 31, 105 S.Ct. at 1711-12 (O’Connor, J., dissenting) (criticizing the majority’s failure “to limit its holding to the use of firearms” only); see also Mathis v. Parks, 741 F.Supp. 567, 572 (E.D.N.C.1990) (“Gamer makes clear that there are limits to the use of deadly force against arrestees, regardless of the nature of the instrumentality used in applying the force.”). Because defendants have failed to produce a scintilla of evidence on the issue of the frequency and severity of the injuries caused by LAPD dogs, I cannot accept my colleagues’ unexplained assumption that LAPD dogs such as Volker are “meaningfully distinguishable” from a gun in *1462that they pose no “substantial risk of causing death or serious bodily harm.” Model Penal Code § 3.11(2).
In sum, after Gamer, no reasonable officer could believe it lawful to authorize the use of deadly force against suspects who do not pose “a significant threat of death or serious physical injury to the officer or others.” Garner, 471 U.S. at 3, 105 S.Ct. at 1697. Accordingly, the individual defendants’ claim of qualified immunity must necessarily turn on the factual question whether a reasonable policymaker could have believed that dogs trained pursuant to the LAPD canine policy did not pose “a substantial risk of causing death or serious bodily harm.” Model Penal Qode § 3.11(2). In their motion for summary judgment, defendants stand mute on this question. They merely ask that their policy, as written, be held constitutional as a matter of law. At this stage of the litigation we cannot cloak the individual defendants with qualified immunity since there is no evidence that would permit us to determine that a reasonable officer could have believed LAPD dogs such as Volker do not constitute deadly force.
CONCLUSION
Defendants are not entitled to summary judgment on the ground that the LAPD canine policy, as written, is constitutional as a matter of law. Furthermore, the City is not entitled to summary judgment on a Monell defense because, for purposes of summary judgment, it admitted to being the moving force behind Officer Bunch’s deployment of Volker. Finally, the individual defendants who are responsible for the LAPD canine policy are not entitled to summary judgment on the basis of qualified immunity.
. I use the term “canine policy” to include both the LAPD policy regarding the dogs’ training as *1452well as its policy governing the officers' use of the dogs.
. See Defendants' Motion for Summary Judgment at 1 ("The policy or procedure sued [sic] by the Los Angeles Police Department to employ police dogs to find and to seize felony suspects who refuse to surrender is not in violation of the Fourth Amendment...id. at 8, 105 S.Ct. at 1699-1700 ("THE POLICY OR PROCEDURE OF THE LOS ANGELES POLICE DEPARTMENT IN ITS USE OF TRAINED POLICE DOGS IS CONSTITUTIONAL”).
. The LAPD canine policy explicitly directs that dogs be trained to "seize by biting a concealed suspect who refuses to surrender even if he offers no physical resistance to the dog.” Defendants' Reply at 2. Pursuant to that policy, dogs are trained to keep their bite-hold on a suspect unless and until they are called off by their handlers. See Deposition of Donn Yarnall at 57, 59-61; Deposition of Daniel Bunch at 26-27; Deposition of Patrick McKinley at 44-46.
. See Model Penal Code § 3.11(2) (1962) (deadly force is "force which the actor uses with the purpose of causing or which he knows to create a substantial risk of causing death or serious bodily harm"); see also Mattis v. Schnarr, 547 F.2d 1007, 1009 n. 2 (8th Cir.1976) (en banc) (using Model Penal Code definition of deadly force); Pruitt v. City of Montgomery, 771 F.2d 1475, 1479 n. 10 (11th Cir.1985) (same); Ryder v. City of Topeka, 814 F.2d 1412, 1416 n. 11 (10th Cir.1987) (same); Robinette v. Barnes, 854 F.2d 909, 912 (6th Cir.1988) (same).
. Chew stresses that how a dog is trained and used is critical to the determination whether it constitutes a deadly instrument. In his papers opposing summary judgment, Chew distinguishes between two different ways of training and using police dogs: a “find and bark” policy, where dogs are trained not to attack and bite a stationary, non-resisting suspect, but instead to bark until the dog's handler arrives, and a “find, bite, and hold” policy such as the LAPD's. See Chew Opposition to Defendants' Motion for Summary Judgment at 5-9. In fact, Chew has documented how, throughout the 1980s, the LAPD vacillated between these two different policies. According to Chew, beginning in 1980, the LAPD trained its police dogs to find, bite, and hold suspects even if the suspects were stationary and not resisting the dog. In 1984, this policy was abandoned in favor of a find and bark policy used by many other law enforcement agencies. In 1988, the LAPD returned to its original find, bite, and hold policy. See id. at 5-6. More recently, it has apparently returned to a find and bark policy. See "LAPD Begins Warnings on Use of Dogs," L.A. Times, Sept. 30, 1992, at B3.
. See, e.g., United States v. Schoenborn, 4 F.3d 1424, 1433 (7th Cir.1993) (whether an object constitutes a dangerous weapon under 18 U.S.C. § 113 is a question of fact that necessarily depends on the particular facts of each case); Robinette v. Barnes, 854 F.2d 909, 912 (6th Cir.1988) ("[M]any law enforcement tools possess the potential for being deadly force.... [A]s any faithful reader of mystery novels can attest, an instrument of death need not be something as obviously lethal as a gun or knife. The ubiquitous 'blunt object’ kills just as effectively.”); United States v. Moore, 846 F.2d 1163, 1166 (8th Cir.1988) ("The question of what constitutes a 'deadly and dangerous weapon' [under 18 U.S.C. § 111] is a question of fact for the jury.").
State cases are in accord. See, e.g., People v. Fuqua, 58 Cal. 245, 247 (1881) ("There may ... be cases in which [a weapon's] character ... depends upon the manner in which it was used, and thus becomes a mixed question of law and fact. In cases of the latter kind the character of the weapon must be left to the determination of the jury, under appropriate instructions.”); People v. White, 212 Cal.App.2d 464, 465, 28 Cal.Rptr. 67 (1963) (holding that whether any particular instrument constitutes a deadly weapon is "at most a mixed question of law and fact to be determined by the jury upon proper instructions”); People v. Tophia, 167 Cal.App.2d 39, 47, 334 P.2d 133 (1959); People v. Russell, 59 Cal.App.2d 660, 665, 139 P.2d 661 (1943) ("[I]n determining whether an instrument, not inherently deadly or dangerous, assumes these characteristics, recourse may be had to the nature of the weapon, the manner of its use, the location on the body of the injuries inflicted and the extent of such injuries.”); People v. Simpson, 134 Cal.App. 646, 651, 25 P.2d 1008 (1933).
.For example, in Marley v. City of Allentown, 774 F.Supp. 343 (E.D.Pa.1991), aff'd without opinion, 961 F.2d 1567 (3d Cir.1992), the jury was instructed to decide whether or not the use of a dog that caused serious injuries to a fleeing suspect constituted the use of "deadly” force. Id. at 346. After the jury returned a verdict for the suspect, the city moved for judgment notwithstanding the verdict on the ground that the deadly force instruction should not have been given because the weapon used was merely a dog. The district court denied the motion. See id.; see also State v. Sinks, 168 Wis.2d 245, 483 *1454N.W.2d 286 (Wis.Ct.App.1992); People v. Nealis, 232 Cal.App.3d Supp. 1, 283 Cal.Rptr. 376 (1991); State v. Bowers, 239 Kan. 417, 721 P.2d 268 (1986); People v. Kay, 121 Mich.App. 438, 328 N.W.2d 424 (1982); Michael v. State, 160 Ga.App. 48, 286 S.E.2d 314 (1981).
In Nealis, 232 Cal.App.3d Supp. 1, 283 Cal.Rptr. 376, the California Court of Appeals affirmed a conviction for assault with a deadly weapon that involved the siccing of a Doberman Pinscher. The court explained first that "whether a specific dog in a given case is a 'deadly weapon or instrument' is ultimately a question of fact for the jury." Id. at 4, 283 Cal.Rptr. 376. After emphasizing that the conclusion "will depend upon the circumstances of each case," the court reasoned that there was sufficient evidence regarding the viciousness of the dog to support defendant’s conviction for assault with a deadly weapon. Id. at 6, 283 Cal.Rptr. 376.
. In addition, Chew points to deposition testimony of Sergeant Mooring who estimated that suspects suffer bite injuries in over 40 percent of the cases in which police dogs are deployed. Deposition of Mark Mooring at 72-75.
. Even if it is determined on remand that the use of LAPD dogs such as Volker does not constitute the use of deadly force, I agree with Judge Reinhardt that defendants would still not necessarily be entitled to prevail in this action. The objective reasonableness of the City's canine policy' — as written and as applied to the particular facts of this case — would still turn on a genuine issue of material fact. See Opinion of Judge Reinhardt at 1443-44.
. "Where the officer has probable cause to believe that the suspect poses a threat of serious physical harm, either to the officer or to others, it is not constitutionally unreasonable to prevent escape by using deadly force." Garner, 471 U.S. at 11-12, 105 S.Ct. at 1701-02.
. Just as defendants stood mute in their summary judgment papers on the specific issue whether LAPD dogs such as Volker constitute instruments of deadly force, they stood mute on the issue whether Chew posed a threat to others.
. On this issue, Judge Trott engages in appellate factfinding in saying that Chew posed a significant threat of serious bodily harm to others. See Opinion of Judge Trott at 1464-66. Apparently, Judge Trott believes that police have probable cause to believe that anyone with an outstanding felony arrest warrant who flees by jumping fences and by hiding in places such as scrapyards poses such a threat. See id. at 1465, 1470. In support of his factfinding. Judge Trott approvingly quotes the testimony of Sergeant Yamall, who stated: “You cannot assume that the man is not armed. You will die.” Id. at 1464 (citing Deposition of Yamall at 102). Yamall’s testimony is consistent with the LAPD canine policy which specifically instructs officers to assume that all felony suspects are armed. See LAPD Canine Manual at 41.
However, the position of Sergeant Yamall, the LAPD Canine Manual, and Judge Trott is inconsistent with Gamer, which holds that police cannot automatically assume that all fleeing felony suspects pose a significant threat of serious physical harm to others. See Garner, 471 U.S. at 21, 105 S.Ct. at 1706 (holding unconstitutional the use of deadly force against a felony suspect not known to be armed who fled the scene of a burglary by scaling a fence). Whether Judge Trott personally believes this Supreme Court law is unsympathetic to the realities of modern law enforcement is irrelevant to our disposition of this appeal.
. On this point, Judge Trott's reasoning suffers a breakdown in logic. He fallaciously assumes that just because the use of police dogs is not always unreasonable, the use of police dogs is always reasonable. Opinion of Judge Trott at 1472.
. I also agree with Judge Reinhardt that the verdict against Officer Bunch does not moot or preclude Chew’s claims against the remaining defendants. See Opinion of Judge Reinhardt at 1437, 1438-39.
. On remand, it is inevitable that the City will repudiate the stipulation it made for summary judgment purposes that Officer Bunch released Volker pursuant to LAPD policy. They have already tried to do so on appeal. The City’s strategy makes the deadly force issue critically important to the proper resolution of Chew’s claims. Under the Gamer deadly force analysis, if the LAPD dogs are found to be instruments of deadly force, then the policy itself, insofar as it authorized the use of deadly force against non-dangerous felony suspects, would be unconstitutional. In that event, Officer Bunch's conduct would have been authorized by an unconstitutional canine policy.
.Although Judge Reinhardt and Judge Trott constitute a majority on the qualified immunity issue, I refer to the majority opinion as the “Opinion of Judge Reinhardt” for clarity.
. It is important to realize that as policymakers, the defendants must be held to a stricter standard of accountability than officers forced to make snap judgments in the heat of battle on today’s dangerous streets. We are not addressing here the qualified immunity of a handler who in a split-second decision releases his dog against a suspect who shows a glint of metal in his hand. Instead, we are addressing the qualified immunity of high-level city officials entrusted with the weighty responsibility to study, analyze, draft, and implement an effective police policy that does not violate the Constitution.
. "Although we cannot ignore the fact that, in this case, the use of a police dog did result in a person's death, we also cannot ignore the evidence in the record which indicates that this tragic event was an extreme aberration from the outcome intended or expected.” Id.
. Compare Opinion of Judge Reinhardt at 1448-49 ("Here, we must determine whether it was clearly established that it was unlawful to use police dogs to search for and apprehend concealed suspects by biting and seizing them.") (emphasis added) with Mendoza, 27 F.3d 1357, 1361-62 (“It is clearly established that '[t]he use of excessive force by police officers in an arrest violates the arrestee's Fourth Amendment right....' This analysis applies to any arrest situation where force is used, whether it involves physical restraint, use of a baton, use of a gun, or use of a dog. We do not believe that a more particularized expression of the law is neces*1461sary....") (citation omitted) and Marley, 774 F.Supp. at 345-46 (the officer “should have been aware of the constitutional constraints enunciated in Gamer, and it was not objectively reasonable for him to think that unleashing a trained attack dog" to apprehend this fleeing suspect comported with Gamer).
. My colleagues miss the point when they say that Marley could not have guided the Chew defendants in adopting the LAPD canine policy because that case was decided after the incident in this case. I do not cite Marley for that purpose; I cite it for its careful analysis of a similar plea of qualified immunity based on Robinette — first, that Marley considered generic excessive force law from cases such as Gamer; second, that Marley recognized that particular dogs could be instrumentalities of deadly force and that that fact question is one for the jury; and third, that Marley discounted the value of the fact-specific case of Robinette in its qualified immunity analysis.