dissenting, in which HATCHETT, Chief Judge, joins:
The issue before the court today is not an easy one. Prisons certainly have the obligation to prevent the spread of the HIV virus among inmates. At the same time, they have the obligation to protect the rights of prisoners already infected with the virus. This difficult balancing act requires sensitivity to the rights of all prisoners within the state’s custody — both those infected with the disease and those who are not. By upholding on this record Alabama’s policy of excluding HIV 4- inmates from all the programs and activities offered to inmates housed in the general population, the majority permits Alabama to forego any consideration of the rights of HIV+ inmates even where the rights of uninfected inmates are not, in any real sense, endangered. The majority upholds the blanket exclusion on the grounds that any cognizable risk of HIV transmission, no matter how infinitesimal and even if based on a wholly unlikely and speculative chain of events, suffices to disqualify HIV+ inmates from participating in all general population prison programs and that no reasonable accommodation could eliminate this risk. The majority’s holding is in direct conflict with governing Supreme Court precedent and eviscerates § 504’s protections. Moreover, in affirming the judgment of the district court, the Court places its imprimatur on the district court’s failure to conduct a discrete analysis of the risk of HIV transmission in each of the many programs and activities at issue here in derogation of the specific mandate of this court. As a result, HIV+ inmates in Alabama are excluded from participating in a whole host of programs to which uninfected inmates routinely have access — programs and activities which comprise virtually all features of institutional life — without any meaningful inquiry into whether this stigmatizing exclusion is necessary to protect uninfected inmates from a significant risk of HIV transmission. For these reasons, I respectfully dissent.
In School Board of Nassau County, Fla. v. Arline, 480 U.S. 273, 107 S.Ct. 1123, 94 L.Ed.2d 307 (1987), the Supreme Court made clear that recipients of federal aid *1306are not entitled to demand absolute segregation from individuals with contagious diseases; § 504 prohibits those receiving federal moneys from unduly indulging their fears and requires them to tolerate less than significant risks of transmission. In Arline, the Supreme Court held that a school teacher who was diagnosed with tuberculosis was a handicapped individual within the meaning of § 504 and remanded the case for an “individualized inquiry” as to whether Arline could serve as a teacher without posing a significant risk of communicating her disease to her students. Id. at 287, 107 S.Ct. 1123. Observing that “[f]ew aspects of a handicap give rise to the same level of public fear and misapprehension as contagiousness,” the Court explained that § 504 was “carefully structured to replace such reflexive reactions ... with actions based on reasoned and medically sound judgments.... The fact that some persons who have contagious diseases may pose a serious health threat to others under certain circumstances does not justify excluding from the coverage of the Act all persons with actual or perceived contagious diseases.” Id. at 284-85, 107 S.Ct. 1123 (emphasis in original). Accordingly, the Court held that an individualized inquiry into risk was “essential if § 504 is to achieve its goal of protecting handicapped individuals from deprivations based on prejudice, stereotypes, or unfounded fear, while giving appropriate weight to such legitimate concerns of grantees as avoiding exposing others to significant health and safety risks.” Id. at 287, 107 S.Ct. 1123.
In formulating the “significant risk” standard, the Court explained that “[a] person who poses a significant risk of communicating an infectious disease to others in the workplace will not be otherwise qualified for his or her job if reasonable accommodation will not eliminate that risk.” Id. at 287 n. 16, 107 S.Ct. 1123. The Court instructed lower courts to look to objective medical evidence concerning four factors in determining whether a risk is significant: “(a) the nature of the risk (how the disease is transmitted), (b) the duration of the risk (how long is the carrier infectious), (c) the severity of the risk (what is the potential harm to third parties) and (d) the probabilities the disease will be transmitted and will cause varying degrees of harm.” Id. at 288, 107 S.Ct. 1123.
Focusing exclusively on Arline’s third factor, the majority holds that, because of its deadly consequences, any cognizable risk of HIV transmission is, as a matter of law, a significant risk. The majority purports to apply Arline, stating that “the significance of a risk is a product of the odds that transmission will occur and the severity of the consequences,” Maj. op. at 1297, but then proceeds to ignore all but the severity of risk factor. The fatal consequences of a contagious disease, in the majority’s view, suffice to render a transmission risk significant even if the probabilities of transmission are so low as to approach zero, so long as transmission could theoretically occur, letting one factor overwhelm the entire Arline analysis. Although it disclaims any intent of establishing an “any risk” standard, in fact the majority opinion does exactly that. By focusing only on possibilities' — whether “a certain event can occur” and whether “the event can transmit the disease,” Maj. op. at 1299 (emphasis added) — the majority requires a plaintiff to prove that transmission is impossible. This reasoning conflicts with Crime’s explicit directive to consider other relevant factors and is directly contrary to the Supreme Court’s recent decision in Bragdon v. Abbott, 524 U.S. 624, 118 S.Ct. 2196, 141 L.Ed.2d 540 (1998).
In Bragdon, the Court considered whether a dentist could refuse to treat an HIV + patient based on the risk of transmission of the HIV virus. It bears noting that the Supreme Court specifically rejected the very argument the majority adopts today. Bragdon declined to permit discriminatory treatment of those afflicted with the HIV virus based on an “any risk” standard, explaining that “[bjecause few, if *1307any, activities in life are risk free, Arline and the ADA do not ask whether a risk exists, but whether it is significant.” Id. at 2210. Thus, the “significant risk” analysis begins with the existence of a risk of transmission, but it does not end there. “The question under the statute,” the Court stated, was “one of statistical likelihood” based on an “objective assessment of the risks posed.... ” Id. at 2211-12. Indeed, in Bragdon, the Court rejected the dissent’s argument that a significant risk of HIV transmission existed based on instances of possible transmission of HIV from patients to dental and other healthcare workers, finding that this evidence of possible transmission was insufficient to establish “the objective, scientific basis for finding a significant risk to the petitioner.” Id. at 2212. See also Abbott v. Bragdon, 163 F.3d 87, 89-90 (1st Cir.1998) (reaffirming conclusion that seven cases of possible transmission were insufficient to establish significant risk). The majority wholly ignores these principles, failing to recognize Bragdon’s distinction between “any risk” of HIV transmission and a “significant risk,” and offers no reason why Bragdon should not be applied here.
This Circuit has, until today, declined to sanction segregation of those with HIV or AIDS based on an “any risk” standard. In Martinez v. School Bd. of Hillsborough County, Fla., 861 F.2d 1502 (11th Cir.1988), a case summarily dismissed without any analysis by the majority, we considered whether, consistent with the dictates of § 504, a mentally retarded girl with AIDS could be provided schooling in a segregated setting to avoid the transmission of her disease to her classmates. Applying Arline, we rejected the district court’s conclusion that a “remote theoretical possibility” of transmission through tears, saliva, and urine were sufficient to support segregated schooling. This possibility, we explained, “does not rise to the ‘significant’ risk level that is required for Eliana to be excluded from the regular ... classroom.” Id. at 1506. Despite the lethal consequences of the plaintiffs AIDS, we found that the risk of transmission occurring in this manner was too remote and speculative to constitute a “significant risk” under Arline. See also Doe v. Dekalb County School Dist., 145 F.3d 1441, 1446 (11th Cir.1998) (remanding to permit the district court to “explain why it believes that the risk posed by Doe is ‘remote’ ” under Martinez and to “make findings of fact with respect to the Arline factors”).
I recognize that sex and needle sharing have been established as pathways of HIV transmission. However, different programs present different levels of risk of such behavior occurring. The prior mandate of this court recognized this distinction by requiring that the district court evaluate each program in terms of the risks involved. For example, the risks of residential integration differ from the risks of integrating a religious service or an educational class. Even where an established- transmission pathway is involved, a risk of HIV transmission must be evaluated on objective evidence of the likelihood of transmission and the other Arline factors, not on mere speculation. Contrary to the majority’s suggestion, this standard does not demand that someone die before a risk is considered significant. It does, however, demand both evidence that particular conduct will transmit the disease and a reasonable likelihood that the conduct will actually take place in the particular program at issue. See Estate of Mauro v. Borgess Medical Ctr., 137 F.3d 398, 402-03 (6th Cir.) (because “neither [the Rehabilitation Act nor the ADA] requires the elimination of all risk posed by a person with a contagious disease,... our analysis ... must not consider the possibility of HIV transmission, but rather focus on the probability of transmission weighed with the other three factors of the Arline test”) (emphasis in original), cert. denied, — U.S. -, 119 S.Ct. 51, 142 L.Ed.2d 39 (1998); id. at 409 (Boggs, J., dissenting) (“[T]he ‘significance’ of risk inescapably involves a judgment about the probabilities of harm.”); Chalk v. United States Dist. Ct., 840 F.2d 701, 709 (9th Cir.1988) (holding that “it was error to *1308require that every theoretical possibility of [HIV transmission] be disproved”); H.R.Rep. No. 101-485, pt. 2 at 56 (1990), reprinted, in 1990 U.S.C.C.A.N. 267, 338 (noting that “a speculative or remote risk” is insufficient to support a finding of a “significant risk”); id., pt. 3 at 46, reprinted in 1990 U.S.C.C.A.N. at 469 (“The plaintiff is not required to prove that he or she poses no risk.”). It is one thing to say that objective evidence of a small risk of transmission of a deadly, contagious disease entails a significant risk in a particular context. It is quite another to say, as the majority does, that the probability of transmission is irrelevant so long as transmission is theoretically possible.
In this case, the district court asked only whether there was a risk of HIV transmission, equating a mere possibility of HIV transmission with the significant risk required for exclusion. No separate analysis of each program weighing the different risks involved and the likelihood of transmission was conducted to determine whether plaintiffs were otherwise qualified to participate in any of them. In program after program, the district court concluded that integration would pose a significant risk, relying on and repeating verbatim the following paragraph thirty-nine times in the course of its opinion to justify its holding that the HIV + inmates were not “otherwise qualified” to participate in any general population programs:
The Court of Appeals recognized in Harris, 941 F.2d at 1495, 1519-20 that “high risk behavior occurs disproportionately in prison systems” and that “under any system of prison administration, the elimination of high risk behavior, such as homosexual behavior or IV drug use is impossible.” This Court, too, recognizes that elimination of high risk behavior is impossible. There is no guarantee that any conduct will or will not occur in a prison setting. This Court is ill-suited to instruct prison officials on the likelihood of the occurrence of high risk behavior. This Court is even more ill-suited to instruct the prison system on when and how to prevent such conduct when they, along with their managers and medical officials, determine that such conduct is likely. Because the Defendant/Prison system has decided that such conduct is likely, and because of the catastrophic severity of the consequences if such conduct does occur, this Court holds that integra-ti[o]n[ ] ... would present a significant risk of transmitting the deadly HIV virus.
Op. at 62-63 (footnotes omitted) (religious programs).1
Instead of following this court’s mandate in Harris v. Thigpen, 941 F.2d 1495 (11th Cir.1991), requiring findings concerning “the risk of transmission ... with regard to each program from which appellants have been automatically excluded,” id. at 1526, the district court simply deferred to the unsubstantiated suspicions of prison officials that transmission would occur.2 *1309Even where there had been no history of high-risk activity in connection with a particular program, the district court attached no significance to the absence of such activity, noting that “the lack of incident reports ... does not necessarily prove that high risk behavior does not occur,” Op. at 61, only that such behavior “has not been discovered.” Op. at 62. In its analysis of these programs and activities, the district court assumed that high-risk behavior would eventually occur, explaining time and again that prison officials must expect the unexpected, thereby “allowing] the DOC to ... guard against contingencies which, under normal circumstances, would be considered only remote possibilities.” Op. at 58. This rejection of evidence in favor of sheer speculation is reflected, for example, in the district court’s discussion of the out-of-prison programs.3 Although acknowledging the evidence that “surveillance by a correctional officer would likely prevent [high-risk] behavior,” Op. at 152, in some of these programs, the district court justified its conclusion by constructing the following hypothetical:
[I]n the prison system, one must always be prepared for the unexpected. An automobile or other accident may incapacitate a guard and leave the inmates on the out-of-prison detail free to proceed without an escort. An inmate, temporarily healthy but facing the bleak future of all sero-positives, may have controlling impulses vastly different from those of healthy inmates who may be more concerned about the penal aspects of their future....
If ... a correctional officer has been incapacitated in an accident, a female inmate with contempt for that officer could purposely implant, while the guard remains unconscious, blood containing HIV organisms from the HIV + inmates open wounds. An unconscious guard would be defenseless, and would perhaps never be fully aware of how the infection was contracted. The unexpected must be expected in a penal setting or when dealing with convicted felons.
Op. at 158 (footnotes omitted).
As this passage illustrates, the district court’s analysis found that a mere possibility of a transmission risk, often based on nothing more than highly speculative scenarios, justifies the wholesale segregation and exclusion of HIV + inmates from prison programs and activities. Under this reasoning, which the majority approves, an individual is, in essence, subject to segregation and discriminatory treatment simply for having the HIV virus. This is precisely the result that § 504 and Arline’s “significant risk” standard sought to prevent.
Nor are the cases cited by the majority to support its refusal to follow Bragdon applicable. These cases uphold the authority of hospitals to prohibit HIV+ physicians and surgical technicians from performing certain invasive surgical procedures based on public health guidance authority absent here. See Mauro, 137 F.3d 398 (6th Cir.1998); Doe v. University of Maryland Medical Sys. Corp., 50 F.3d 1261 (4th Cir.1995); Bradley v. University of Texas M.D. Anderson Cancer Ctr., 3 F.3d 922 (5th Cir.1993). First, these cases were decided before Bragdon explained the distinction between a risk of HIV transmission and a “significant risk.” Second, none of these cases hold that any risk of HIV transmission- — no matter how small or remote- — -is a significant risk as a matter of law. In Mauro, Doe, and Bradley, the courts explicitly relied on the guidance of public health officials to support their finding of a significant risk and Bragdon teaches us that “the views of public health authorities,” such as the Centers for Disease Control (“CDC”), “are of special weight and authority.” Bragdon, 118 S.Ct. at 2211. In Doe, for example, the court emphasized that the CDC had stated that “hospitals may bar HIV-positive surgeons from performing *1310those procedures identified by the hospital as exposure prone” and the university-concluded that “all neurosurgical procedures that would be performed by Dr. Doe fit the definition of exposure-prone procedures.... ” Doe, 50 F.3d at 1266 (emphasis in original); see also Mauro, 137 F.3d at 404 (“deferring] to the medical judgment expressed in the Report of the Centers for Disease Control”); Bradley, 3 F.3d at 924 (relying on CDC guidelines). These cases, therefore, conclude, based on explicit CDC guidance, that a small risk of transmission in that particular context amounts to a “significant risk.” The record below does not contain any such guidelines or similar evidence. Moreover, the only testimony below in support of Alabama’s policy consists of the opinions of corrections officials and none comes from any public health entity unconnected with the parties.
I likewise believe the majority errs in holding that plaintiffs were not “otherwise qualified” within the meaning of § 504 because of the prison’s penological interests in excluding them from the many programs and activities at issue here. In Hams, our mandate made clear that the prison’s penological interests supporting the segregation of HIV+ prisoners were not automatically sufficient to render plaintiffs unqualified, thereby declining to insulate the Alabama prison authorities from liability under the Rehabilitation Act. We explained that
it is not enough for the district court simply to rely on general findings and prison policy reasons that support segregation .... We ... do not believe ... that the prison’s choice of blanket segregation should alone insulate the DOC from its affirmative obligation under the Act to pursue and implement such alternative, reasonable accommodations as are possible for HIV-positive prisoners ■with respect to various programs and activities that are available to the prison populations at large.
Harris, 941 F.2d at 1527 (footnote omitted).
Under § 504, the “otherwise qualified” analysis focuses on the requirements of the program at issue and whether reasonable accommodations would permit the plaintiff to participate, see Arline, 480 U.S. at 287 n. 17, 107 S.Ct. 1123, not whether the discriminatory exclusion is supported by legitimate goals, penological or otherwise. As Harris makes clear, a § 504 plaintiff is not rendered unqualified to participate in a prison program simply because the prison asserts general penological interests in excluding him or her from participation.
This is not to say that penological concerns have no place in a § 504 analysis. As the Seventh Circuit has explained, “[t]erms like ‘reasonable’ and ‘undue’ are relative to circumstances, and the circumstances of a prison are different from those of a school, an office, or a factory,” and “[t]he security concerns that the defendant rightly emphasizes ... are highly relevant to determining the feasibility of the accommodations that disabled prisoners need in order to have access to desired programs and services.” Crawford v. Indiana Dep’t of Corrections, 115 F.3d 481, 487 (7th Cir.1997). As Crawford suggests, however, a prison’s legitimate peno-logical interests must be considered within the established § 504 framework — not by supplanting this framework with the inapplicable constitutional analysis set forth in Turner v. Safley, 482 U.S. 78, 107 S.Ct. 2254, 96 L.Ed.2d 64 (1987), as the district court did here. Cf. Dothard v. Rawlinson, 433 U.S. 321, 334-37, 97 S.Ct. 2720, 53 L.Ed.2d 786 (1977) (considering prison’s penological interests with in established Title VII framework, not constitutional framework).
Finally, I disagree with the majority’s holding that the district court correctly concluded that plaintiffs had failed to offer any reasonable accommodations, for several reasons.4 First, the majority errs in *1311affirming the district court’s conclusion that hiring any additional officers to reduce the risk of HIV transmission would impose an undue burden on the Alabama prison authorities to provide reasonable accommodation. Distancing itself from the district court’s statement that hiring even a single officer would impose an undue burden, the majority unfairly characterizes the relief requested by the plaintiffs as integration of all the programs and activities challenged here, which it asserts would require hiring many additional officers. As the majority must concede, the inability to accommodate plaintiffs in all the programs and activities would not justify Alabama’s refusal to provide accommodation in any of the programs and activities at issue here. The majority nonetheless affirms the district court because plaintiffs did not specify which programs and activities they would prefer to integrate. This begs the question since the district court failed to conduct the correct inquiry into each program as .to whether integration would pose a significant risk and what measures would be required to accommodate plaintiffs. Thus, this question should also be remanded for a proper evidentiary hearing.
Second, by rejecting the use of the prison risk classification system as a reasonable accommodation because it is not error-free, the majority once again requires a plaintiff to prove that he or she can not possibly transmit the HIV virus, thereby equating any possible transmission risk with the significant risk required to support exclusion under Arline and Bragdon. It may well be that the prison risk classification system will not suffice to reduce the risk of transmission resulting from integration to a less than significant one. At this point, however, the record is inadequate to establish the rate of error in the classification program in accordance with Arline’s dictates. The district court excluded much, though not all, of the evidence about this classification scheme and the district court only found that this system was not error-free. It made no findings regarding the likelihood of error. Accordingly, I believe that the district court’s exclusion of evidence on this point requires a new trial.
I would note that it is significant that Alabama prison authorities use this selection process despite the possibility of error in the risk classification system and despite the tragic, and even fatal, consequences of error, which have in fact occurred. The majority offers no explanation why the prison system should be entitled to insist its risk classification system work in an error-free manner in the HIV context, and this context alone. Further, it appears that Alabama does not impose a similar policy of segregation and exclusion where other inmates have other contagious diseases, such as hepatitis.
Congress and the Supreme Court have recognized that automatic expulsion and exclusion from society because of “health risks” have historically masked discrimination against individuals with contagious diseases based on prejudice and baseless fear. Certainly, care must be taken to minimize risk of infection, but care must just as certainly be taken to assure that a society that aspires to be just does not make outcasts of its stricken citizens. In this case, no fair assessment of the risks involved was conducted and no fair consideration was given to reasonable accommodations which could have minimized to an acceptable level those risks which did exist. No balancing of rights occurred here. Thus, I respectfully dissent.
. This same passage was recited throughout the district court’s opinion. See Op. at 72-73 (rehabilitation programs); 82-83 (visitation); 111-12 (vocational programs); 123 (recreational programs); 132-33 (dining hall); 141-42 (medical clinic); 155-56 (out-of-prison programs); 165 (educational programs); 178-79 (Laubach literacy program) 187 (sewing factory); 195-96 (data processing jobs); 203-04 (kitchen jobs); 211-12 (yard workers); 221-22 (laundry jobs); 230 (trash recycling jobs); 238-39 (health care uniL jobs); 246-47 (gate runner jobs); 254-55 (hair cutting jobs); 262-63 (use of library/library jobs); 277-78 (Double 0 Squad jobs); 285 (tractor operator jobs); 296-97 (maintenance jobs); 304 (trash detail jobs); 313-14 (runner jobs); 321-22 (paralegal training class); 331-32 (educational programs); 344 (graduation ceremonies); 363-64 (vocational programs); 378-79 (hair cutting jobs); 386-87 (laundry jobs); 394 (garden jobs); 402-03 (visitation); 410-11 (medical clinic visits); 420-21 (rehabilitation programs); 429-30 (recreational programs); 439 (library); 446-47 (religious programs); 453 (Alabama Volunteers in Correction program).
. The majority, too, makes this error, affirming the district court based on evidence of high-risk behavior in the prison system as a whole, not in each program as Harris required.
. These programs permit inmates to leave prison, generally in handcuffs, under direct surveillance of prison guards, to talk to school children about the perils of drug use, to attend a funeral or seek medical care, or to work outside the institution fence.
. I would note that both the Supreme Court and this Court have repeatedly recognized that § 504 requires covered recipients to make reasonable accommodations to permit individuals with disabilities to participate in *1311programs offered by a recipient of federal financial assistance. See Alexander v. Choate, 469 U.S. 287, 300-01, 105 S.Ct. 712, 83 L.Ed.2d 661 (1985); Harris, 941 F.2d at 1525-27; Martinez, 861 F.2d at 1505-07. Thus, I find no basis for the majority’s suggestion that § 504 imposes no reasonable accommodation mandate and that the source of any duty to provide reasonable accommodations comes solely from the applicable agency regulations.