dissenting:
Put yourself in the shoes of Sergeant Bill Nulton of the San Diego Police Department. Late one Thursday night in *857February, around 11:00 pm, you are dispatched to respond to a burglar alarm that has gone off at a two-story commercial building. You arrive at the scene within two or three minutes of getting the call, together with your police service dog,Bak, and two other officers. Approaching the building, you do not see anyone leaving the building or the parking lot. You inspect the building and see that two doors on the second floor are open; You go to the second floor-and determine that' one open door leads to a bathroom, which is empty. Another door is closed and locked. The remaining door leads to Suite 201. It is propped open. The building is dark. You cannot see inside and do not know whether anyone is there. You yell loudly, “This is the San Diego Police Department! Come out now or I’m sending in a police dog! You may be bitten!” There- is no response. You wait- between 30 and 60 seconds, but still no response. You repeat the same warning one or two more times. Again, no response. ■ Because nobody has responded to the warnings, you are concerned that if there is someone inside the building who triggered the alarm, that person may be a burglar lying in wait. You have no way of knowing whether that person is armed. What would you do?
Unfortunately for Sgt. Nulton and for all law enforcement officers within the Ninth Circuit, if you release your trained service dog and follow him with a flashlight to search for a suspect, you might wind up in trial. Thanks to the majority opinion, officers will be discouraged from protecting themselves and encouraged to risk their lives by exposing themselves to any burglar who might be armed and lying in wait, either because they cannot use a dog at all or must remain so closely "tethered to the dog that they necessarily have to expose themselves to the potentially armed burglar.1
•I respectfully dissent, .
The majority opinion dutifully and accurately recites, at 847, that we are supposed to evaluate the reasonableness of the force used “from the perspective of a reasonable officer on the scene, rather than with the 20/20 vjsion of. hindsight.” Graham v. Connor, 490 U.S. 386, 396, 109 S.Ct. 1865 (1989). That is the law, but that is not what the majority actually. doe.s.
Consider the attention given in the majority opinión to the testimony of Sara Lowry that she did not hear the warnings yelled loudly and more than once by Sgt. Nulton. See majority op. at 851-52 & 853. I accept that to have been true; she was asleep on, the couch after "consuming five vodka drinks earlier in the evening. But that’s not something that the officer could have known. The majority explicitly states, at 853, that the fact that Lowry did not hear the warnings diminishes the weight it is willing to give to the fact that the officer gave warnings. That is not an evaluation based on the perspective of a reasonable officer on the scene.
Similarly, the majority opinion emphasizes, at 849, that “Lowry did not pose a threat to the officers or others.” See also majority op. at 849-51. But the officers did not know and could 'not have 'known that Lowry would be the one and only person they would encounter inside the building. The majority opinion rests on “the 20/20 vision of hindsight.” Thát is error.
*858When viewed from the appropriate perspective, there is no interpretation of the facts that could lead a jury to conclude that the use of a police dog under the circumstances was excessive force. The summary judgment entered by the district court in favor of the City should be affirmed.
1. Facts
The facts, from the perspective of the officers, were as follows. At approximately 10:40 pm on the night of Thursday, February 11, 2010, a burglar alarm went off in a two-story office building in San Diego. Three police officers arrived within minutes of dispatch to find a darkened building and, on the second-story balcony, two open doors, including one that was propped open.2 After scaling the ground-floor gate, the officers looked around the second story for any other indication that the building was occupied by someone who belonged there, finding none. The officers determined that one of the open doors led to an empty bathroom. They announced themselves loudly, more than once, at the door that was propped open.3 No one responded.
At this point, the officers were forced to make a decision, aided only by the information above and a handful of commonsense assumptions. Doors do not generally open of their own accord, particularly in an empty office building. Nor do people often come to an office for a legitimate purpose at nearly 11 o’clock at night without turning on the lights. Those who do are unlikely to fail to notice if they set off a burglar alarm, especially as it was loud enough to be heard from the parking lot. And there is no obvious reason why someone who was not actively trying to hide from the police would fail to respond to a command to exit a building. In other words, the totality of the circumstances strongly suggested that if there was- someone in the building, that person was likely a burglar, and possibly armed.
II, The use of force was reasonable
The question in an excessive force analysis is “whether the officers’ actions are .‘objectively reasonable’ in light of the facts and circumstances confronting them.” Graham, 490 U.S. at 397, 109 S.Ct. 1865. We assess reasonableness by looking at three factors: (1) “the type and amount of force inflicted,” (2) “the government’s interest in the use of force,” and (3) the balance between “the gravity of the intrusion on the individual” and “the government’s need for that intrusion.” Glenn v. Washington Cty., 673 F.3d 864, 871 (9th Cir.2011) (internal citations omitted). The majority ignores Graham’s admonition that courts must consider reasonableness from the perspective of the officers under the circumstances and instead veers between viewing the facts from Lowry’s perspective, to the exclusion of the officers’, and speculating on what could have happened under different circumstances. Viewed from the appropriate perspective, all three factors of the analysis under Graham weigh in favor of the officers.
A. The type and amount of force inflicted was not severe
The majority begins its analysis by reading into our ease law a blanket rule that has never existed before. We have never held that the use of a police dog is categor*859ically “severe,” as the majority opinion suggests, at 847. Rather, the - cases the majority cites all involved an individual analysis of the use of force under particularized circumstances.
In Smith v. City of Hemet, it was the Hemet Police Department itself that categorized the use of a police service dog as “intermediate” force, which by its description was “the most severe force authorized short of deadly force.” 394 F.3d 689, 701-02 (9th Cir.2005) (en banc). That usage suggests only that intermediate force was the name used by that department for the level of force between light and deadly. It is not a holding that requires us to consider the use of a police dog severe in every case. To the contrary, our opinion in Smith went on to consider the facts in that case— facts very different from those here. The officers in that case sicced a police dog on Smith three times, including once after he had already been pinned down, and then pepper sprayed his open wounds. Id. at 702. Similarly, in Chew v. Gates, the court held that “the force used to arrest Chew was severe” because the dog bit Chew three times, dragged him between four and ten feet, and “nearly severed” his arm. 27 F.3d 1432, 1441 (9th Cir.1994). Meanwhile, in Miller v. Clark County, this circuit found the use of force reasonable when a fleeing suspect suffered a dog bite that lasted between forty-five and sixty seconds, “shredded” his muscles, and went down to the bone. 340 F.3d 959, 961 (9th Cir.2003).
Applying the ease-by-case analysis employed in our previous police dog cases, the district court here properly concluded that the force used against Lowry was “moderate.” In sharp contrast to the grisly injuries in the cases cited by the majority, Lowry required only three stitches in her upper lip, and experienced no visible scarring. The contact was so brief that Sgt Nulton did not even know if contact had occurred. Even Lowry described it as “very quick.” Moreover, although, you might not realize it-from the description in-the majority opinion, Nulton did not simply let Bak go-but rather followed closely behind her as she cleared the small office suite. The district court took this into consideration, noting that “Sergeant Nul-ton was present and immediately called the dog off upon seeing Plaintiff on the couch.”
The facts of this case undermine the majority’s conclusion, at 848-49, that the type of force used against Lowry was comparable to that in Chew. In that case, the police dog was, “beyond the reach of a countermanding order” when it found the plaintiff, and dragged him at least four feet and possibly as many as ten feet before releasing him. Chew, 27 F.3d at 1441. “[T]he longer- a dog is permitted to bite a suspect, the greater the likelihood the suspect will be injured severely.” Miller, 340 F.3d at 963. Here, the district court properly recognized that the risk inherent in the officers’ use of. force was significantly lessened by Sergeant Nulton’s close proximity to Bak and his ability to call her off within mere moments of contact. These circumstances cannot support a conclusion that either the type or the amount of force used was comparable to the “severe” force used in cases like Chew and Smith.
B. The city’s interest in the use of force ■ was strong
On the second step of the Graham analysis, which asks the court to consider the City’s interest in the use of force, ■ the majority errs by viewing the circumstances from Lowry’s-perspective. Generally, this step requires the court to assess “(1) the severity of the crime at issue, (2) whether the suspect posed an immediate threat to the safety of the officers or others, and (3) *860whether the suspect was actively resisting arrest or attempting to evade arrest by flight,” considered in the totality of the circumstances. Miller, 340 F.3d at 964. When viewed from the appropriate perspective, all of these considerations weigh in favor of the City..
1. The officers reasonably anticipated a potential threat
I begin, as does the majority, with an analysis of whether Lowry presented an immediate threat to the safety of the officers. It is irrelevant to this inquiry that Lowry “remained fast asleep on the couch” during the officers’ search' of the building.' See majority op! at 849. Rather, the facts must be viewed from the perspective of the officers, who knew only that they had been called to a building showing signs of a break-in,4 that the building was dark and that as a result any entry would be blind, and that it was nevertheless their job to enter and investigate.
Our precedents make it clear that, when confronted with signs of a burglary/ investigating officers are entitled to protect their own safety. “[Bjurglary and attempted burglary are considered to carry an inherent risk of violence.” Sandoval v. Las Vegas Metro. Police Dep’t, 756 F.3d 1154, 1163 (9th Cir.2014). “Normally, when officers suspect a burglary in progress, they have no idea who might be inside and may reasonably assume that the suspects will, if confronted, flee or offer armed resistance.” Frunz v. City of Tacoma, 468 F.3d 1141, 1145 (9th Cir.2006). “So long as the officers have established probable cause for a burglary, ‘[i]n such exigent circumstances, the police are entitled to enter immediately, using all appropriate force.’ ” Sandoval, 756 F.3d at 1163 (quoting Frunz, 468 F.3d at 1145). While it is true, as the majority observes, that the legal issue in Frunz concerned a Fourth Amendment claim of warrantless entry rather than a claim of excessive force, that distinction does not contradict the factual truth of our observation in Frunz. Neither does it affect the analysis as to the City’s interest in the use of force, where the concern is not with the nature of the force used but rather whether the suspect posed a threat’ to the officers. The factual reality of this threat is not dependent on the point of law at issue.
Here, the officers reasonably suspected that a burglary had taken or was taking place., They made this informed judgment not simply because a burglar alarm had been tripped, but in the totality of the circumstances, and they took reasonable steps to protect themselves. As the district court found, this factor of the analysis weighs in favor of the City.
2. Lowry did’ not resist or attempt to évade arrest
The second factor of the analysis, whether Lowry was resisting or attempting to evade arrest, weighs heavily in her favor in the majority opinion, at 851-52. It is undisputed that she “took no threatening actions (other than non-compliance with shouted orders)” towards the police. Glenn, 673 F.3d at 875. But the majority" fails again to consider the perspective of the police officers. They did not know *861that she was there, for she hadn’t responded to their warning calls. Neither - did they know that she was the only other person in the building.
Unlike the cases relied upon by the majority opinion, the officers here. did not have the benefit of being able, to see the suspect and make a reasoned judgment as to whether that person was resisting or evading arrest. Instead, they had only the information that .whoever was in the building had failed to. respond to their warnings. They could not know whether the unknown occupant would resist. On the facts of this case, from the perspective of the officers, this factor does not point either way,
3. Burglaiy carries an inherent risk of violence
The final factor in assessing the use of force is the severity of the crime at issue. As noted above, wé have found burglary to “carry an inherent risk of violence,” as recently as 2014. Sandoval, 756 F.3d at 1163. The fact that not all burglaries involve violence does not mean that there is no risk of violence, or that police have any way of telling before it is too late which burglars are violent and which are not.
While the majority minimizes the risks of police work, during the year' in which the events of this case occurred eight police officers were killed and 5,074 were assaulted while investigating suspicious people or circumstances, in addition to the three killed and 899 assaulted during a burglary in progress. Criminal Justice Services Division, Law Enforcement Officers Killed & Assaulted 2010, Tables 19 and 73, FBI, https://www.fbi.gov/about-us/ cjis/ucr/leoka/2010. The majority trumpets that only- seven percent of burglaries nationwide involved incidents of violence, at 851-52, but that means that approximately one in every 14 burglaries involved violence, a significant number for police officers who respond to burglary calls regularly.5 The majority’s unmerited jump from the fact that not all burglaries involve violence to the -conclusion that police should not protect themselves against the possibility that violence may occur minimizes the real danger that -police officers face regularly in an inherently risky job. As the district court properly concluded, this factor weighs in favor of the City:
4. Other relevant considerations weigh in favor of the City
The majority opinion errs agáin in weighing other relevant considerations: (1) whether a warning was given before force was used, and (2) whether there were alternative tacties the officers could have used. - And, it entirely disregards another factor that inerits consideration: the physical setting and the likelihood of encountering an innocent bystander.
As to the first of these considerations, all three officers stated that a warning was given. This factor weighs strongly in favor of the City. See Nelson v. City of Davis, 685 F.3d 867, 882 (9th Cir.2012) (quoting Deorle v. Rutherford, 272 F.3d 1272, 1284 (9th Cir.2001) (“‘[T]he' giving of a warning or the failure to do so is a factor to be considered in applying thé Graham balancing test.’ ”)). As noted above, the majority’s observation that Lowry did not liear the’warning is irrelevant from, the perspective of a reásonable officer on the *862scene. The officers had no way to anticipate the unlikely circumstance that someone would be asleep in a nonresidential building late at night and therefore unable to hear a warning.
The majority opinion goes wrong regarding the warning in a second way. It cites, at 853, our decision in Nelson to support its conclusion to weigh this factor only “slightly” in the officers’ favor, but that authority actually provides it no support. The reason that we concluded in Nelson that the warning should not be given much weight was because the officers issued the warning without amplification into a crowd of some 1,000 partygoers at a distance of 45 to 150 feet. Nelson, 685 F.3d at 872, 874. The officers on the scene were obviously aware of the setting and should have known that they could not expect people to hear a warning shouted at a distance into a rowdy crowd. That was not the setting here.' A warning called into a silent three-room office suite late at night should be more than sufficient to place any occupants on alert. The officers here had no reason to expect otherwise.
The majority opinion similarly overreaches in speculating about the alternative means that the officers could have used to investigate Lowry’s office suite. It asserts that Sgt. Nulton could have kept Bak on the lead so that he could have exercised greater control. See majority op. at 853. But the policy of allowing dogs off-lead is in place to protect officers’ safety, as the officers understood. If Bak were kept on the leash, Sgt. Nulton would have been required to expose himself to whomever might have been lurking in the dark office, possibly armed. Moreover, it is far from clear that the results would have been any different had Bak been kept on the lead. Sgt. Nulton was in the room when Bak jumped on Lowry and was able to call off the dog quickly. Unless the leash was especially short — which would have minimized the utility of having a dog at all — a leash may not have kept Bak any closer to Nulton than she already was.
The majority’s alternative suggestion, at 853 n. 10, that the officers should have used night-vision goggles, is simply puzzling. There is no indication in the record that these officers — or any regular patrol officers — had access to that technology, typically assoeiatéd with military and SWAT teams rather than local police forces. Nor is there any basis for us to conclude that use of that equipment would have been practical or beneficial. Would it have compromised the officer in any other way, such as by cutting down on peripheral vision or hindering movement? Would it have actually permitted the officer to spot a burglar who could be expected to be hiding behind a desk or in a closet? I don’t know, and the majority opinion gives us no reason beyond rank speculation to believe that it knows, either.
The majority fails seriously to consider the specific context that the police officer faced and the relative risks that he had to balance when-he decided to release and follow the dog. Context matters. The majority opinion notes, at 853, for instance, that the San Diego police department’s manual requires an officer to keep a police dog on a lead during a residential- search, “unless the handler can reasonably determine there are no residents or animals in the home.” That makes sense, because in a home the likelihood of encountering an innocent bystander who might be found and injured by the dog is obvious.
The officer involved in this case was not at a home, however. He was at a dark commercial building, late at night, where there had been no response to multiple shouted warnings. That is where he made the decision to release and follow the service dog, and that was not a location where *863an innocent bystander was likely to be found. It was possible that nobody was in the building, but in that case releasing the dog posed no risk of harm because there would not have been anyone to find. The risk of injury from releasing the dog mattered only if there was someone else in the building, and the likelihood of someone else in the building being unable to respond to warning shouts (or the loud burglar alarm) because she was passed out on a couch could not have seemed, from the officer’s perspective, very great. None of that matters to the majority, though. The San Diego police department appreciates that there is a difference between a residence and a commercial building, but the majority opinion does not.
C. The degree offeree used was commensurate with the City’s interest
On the last step of the excessive force inquiry, we weigh the degree of force uséd against the government’s interest in using force. Glenn, 673 F.3d at 871. The discussion above makes clear that the force used here was not severe, and the police had a compelling interest in acting to protect themselves against foreseeable danger in an uncertain situation.
III. There are no material disputes of fact
The majority opinion acknowledges, at 856, that the objective reasonableness of an officer’s actions in the excessive force context, is a “pure question of law” once “the relevant set of facts” has been determined, quoting from Scott v. Harris, 550 U.S. 372, 381 n. 8, 127 S.Ct. 1769 (2007). It asserts that there are questions of fact here that preclude summary judgment, but it does not identify what they are. In reality, they don’t exist. Summary judgment for the officers was appropriate and should be affirmed.
In her briefing, Lowry pointed to several purported factual disputes that she argued should preclude summary judgment: whether the door to her suite was open, whether the suite was dark, and whether the officers provided her with warning pri- or to entering'the suite. If these facts were in dispute, summary judgment might be inappropriate. An open door could be the difference between an overreaction to a false alarm and a reasonable response to an apparent break-in. The illumination in the office is relevant to the officers’ need to protect themselves from a potential ambush. And we have explicitly held that whether a warning was provided is one of the. factors to be considered. See Nelson, 685 F.3d at 882.
However, these facts- are not in dispute, as Lowry presented no admissible evidence on her behalf. The district court excluded Lowry’s' testimony that the door to her office suite was closed because it was not “firsthand”: she did not state that she had actually closed the door but rather relied on a belief that it always closed automatically. Similarly, the district court concluded that her testimony as to the level of illumination in the suite was “entirely speculative.”6 Finally, it concluded that Lowry lacked the proper foundation to testify-as to whether Sergeant Nulton had issued a warning prior to releasing Bak, as she was asleép at the time the warning was given.
The district court’s evidentiary rulings made in the summary judgment context are reviewed for abuse of discretion and can only be reversed if “both ‘manifestly erroneous and prejudicial.’ ” Bias v. Moynihan, 508 F.3d 1212, 1224 (9th Cir.2007) *864(internal citation omitted). Since “[generally, a witness must have ‘personal knowledge of the matter’ to which she-testifies,” Bemis v. Edwards, 45 F.3d 1369, 1373. (9th Cir.1995) (quoting, Fed.R.Evid. 602), it was not manifestly erroneous to conclude that Lowry had no personal knowledge of events that she did not in fact witness. Therefore, we must uphold the district court’s conclusions that the door was open, the suite dark, and the warning given.
■ Without any admissible evidence to suggest that the doors to the office suite were closed, the suite illuminated, or a warning not provided, viewing the facts in the light most favorable to Lowry does not change the analysis above. The majority contends, at 856, that a jury might draw different “inferences” from the facts, but it does not actually say what different factual finding might be made — that the door was not open, that the room was not .dark, or that warnings had not been given. No evidence supports any of those findings. The “inferences” in question amount to the ultimate question of whether the officer’s actions were objectively reasonable. That question, even the majority acknowledges, is a pure question of law. Given .the facts available to the reviewing court, it is clear that the type and amount,of force inflicted was moderate, the City had a strong interest in using force, and the degree of force used was commensurate with the City’s interest in the use of force. As a result, the officers’ actions were constitutional, and there can be no liability under Monell v. Dep’t of Soc. Servs. of the City of N.Y., 436 U.S. 658, 98 S.Ct. 2018 (1978).
I would affirm the district court’s grant of summary judgment.
. Lowry contests that the door was, in fact, ajar. The district court excluded Lowry’s testimony- about the door, however, meaning that she has provided no admissible evidence to support her contention, as' discussed in more detail below, at 863-64.
. Lowry disputes this fact as well, but the district court also excluded the evidence she offered on that subject, as discussed below.
. The majority's focus on the rate of false alarms is misplaced, as well as foolish. See majority op. at 850-51 n. 6 and above at 857 n. 1. The alarm may have been the reason the police arrived on the, scene, but once the officers approached the building^ several other factors, notably the open door and the darkened suite, suggested that something was amiss. For the same reason, the majority's attempt, at 850, to characterize the district court’s finding, for the officers as suggesting that “any person inside an office building where a security alarm has been tripped at night necessarily poses an immediate threat to their safety or that of others” is simply inaccurate.
. In 2011, San Diego police handled a total of 5,840 burglaries;. See Automated Regional Justice Information System, Crime Statistics, http://crimestats.arjis.org/. As of 2014, there were 1,651 SDPD officers on full duty, ’ Melissa Mecija, San Diego Police- Department staffing levels at Ibwest in over a decade, ABC 10 NEWS, (Aug. 12, 2014), http://www.10 news.com/news/san-diego-police-department-staffing-levels-at-lowest-in-over-a-decade-08122014.
, In fact, Lowry herself testified in her deposition that it was "dark” in the suite when she went to sleep, and that there were no lights or computer screens illuminating the room.