concurring in part and dissenting in part:
The majority has written an opinion that is very praiseworthy as far as it goes. I concur in Parts I and II. I also concur in the first portion of Part III, to the end of subdivision (1). Beyond that, I agree substantially with the majority’s discussion leading to the conclusion that the Supreme Court in Lawrence v. Texas, 539 U.S. 558, 123 S.Ct. 2472, 156 L.Ed.2d 508 (2003), applied something more rigorous than traditional rational basis review in striking down Texas’s criminalization of sexual relations between members of the same sex. Finally, I agree that the district court erred in dismissing the complaint for failure to state a substantive due process claim, and that we must remand for further proceedings. Unlike the majority, however, I would also reverse the dismissal of the equal protection claim. But where I differ most from the majority is in the level of scrutiny to be applied to both claims. In my view, the so-called “Don’t Ask, Don’t Tell” statute,1 10 U.S.C. § 654, must be subjected to strict scrutiny. Under that standard, the Air Force must demonstrate that the statute’s restriction of liberty, and its adverse classification of homosexuals, are “narrowly tailored to serve a compelling state interest.” Reno v. Flores, 507 U.S. 292, 301-02, 113 S.Ct. 1439, 123 L.Ed.2d 1 (1993).
Substantive Due Process
As the majority opinion correctly recognizes, the Supreme Court’s opinion in Lawrence never unambiguously states what standard of review it is applying. The Lawrence opinion leaves no doubt at all, however, about the importance of the right it is protecting. In discussing the flaws of Bowers v. Hardwick, 478 U.S. 186, 106 S.Ct. 2841, 92 L.Ed.2d 140 (1986), which it was overruling, Lawrence explained:
To say that the issue in Bowers was simply the right to engage in certain sexual conduct demeans the claim the individual put forward, just as it would demean a married couple were it to be said marriage is simply about the right to have sexual intercourse. The laws involved in Bowers and here are, to be sure, statutes that purport to do no more than prohibit a particular sexual act. Their penalties and purposes, though, have more far-reaching consequences, touching upon the most private human conduct, sexual behavior, and in the most private of places, the home. The statutes do seek to control a personal relationship that, whether or not entitled to formal recognition in the law, is within the liberty of persons to choose without being punished as criminals.
This, as a general rule, should counsel against attempts by the State, or a court, to define the meaning of the relationship or to set its boundaries absent injury to a person or abuse of an institution the law protects. It suffices for us to acknowledge that adults may choose to enter upon this relationship in the confines of their homes and their own *823private lives and still retain their dignity as free persons. When sexuality finds overt expression in intimate conduct with another person, the conduct can be but one element in a personal bond that is more enduring. The liberty protected by the Constitution allows homosexual persons the right to make this choice.
Lawrence, 539 U.S. at 567, 123 S.Ct. 2472 (emphases added). Two points shine forth from this passage and its context in Lawrence: first, the right to choose to engage in private, consensual sexual relations with another adult is a human right of the first order and, second, that right is firmly protected by the substantive guarantee of privacy — autonomy of the Due Process Clause. Thus, even though the Court did not expressly characterize the right as “fundamental,” it certainly treated it as such. It is this treatment, and the important individual values of liberty it recognizes, that require strict scrutiny of governmental encroachment on that right. In my view, therefore, Lawrence itself mandates strict scrutiny of the “Don’t Ask, Don’t Tell” statute.
In order to apply strict scrutiny, however, we do not need to satisfy ourselves that Lawrence commands or expressly adopts that standard of review. We are not reviewing a state criminal conviction, where we are forbidden by the Antiterrorism and Effective Death Penalty Act from applying a constitutional standard unless it has been determined by the Supreme Court. See 28 U.S.C. § 2254(d)(1). In the present context, it is enough that the question is an open one. As the majority opinion recognizes, Lawrence avoids (carefully, it seems) stating what standard of review the Court was applying. Certainly nothing' in Lawrence can reasonably be read as forbidding the application of strict scrutiny to statutes attaching severe consequences to homosexual behavior.2 The question of the standard of scrutiny in this case is therefore an open one, and we must address it according to our best understanding of the individual constitutional rights and governmental action involved.3 For reasons that should already be apparent from my quotation and discussion of Lawrence, I have no difficulty concluding that the right to engage in homosexual relationships and related private sexual conduct is a personal right of a high constitutional order, and that the “Don’t Ask, Don’t Tell” statute so penalizes that relationship and conduct that it must be subjected to strict scrutiny.
Equal Protection
Major Witt presented an equal protection claim to the district court, but acknowledges here that such a claim was rejected by our court in Philips v. Perry, 106 F.3d 1420 (9th Cir.1997). Although she does not pursue it before our three-judge panel, she does preserve her right to assert the claim in the event she seeks en banc review of our decision; she has not *824abandoned the claim.4
I do not believe that Philips ties our hands. Philips applied rational basis review to an equal protection attack on the “Don’t Ask, Don’t Tell” policy of the Navy. It did so on the authority of our earlier decision in High Tech Gays v. Defense Industrial Security Clearance Office, 895 F.2d 563 (9th Cir.1990). See Philips, 106 F.3d at 1425. High Tech Gays, however, was based on the proposition that it would be inappropriate to apply strict scrutiny to classifications targeting homosexuals when the Supreme Court had held in Bowers that homosexual conduct could be made a crime. See High Tech Gays, 895 F.2d at 571 (“[I]f there is no fundamental right to engage in homosexual sodomy ... see [Bowers v.] Hardwick, ... it would be incongruous ... to find a fundamental right of homosexual conduct under the equal protection component of the Due Process Clause of the Fifth Amendment.”). Because Lawrence unequivocally overruled Bowers, it “undercut the theory [and] reasoning underlying” High Tech Gays and Philips “in such a way that the cases are clearly irreconcilable.” Miller v. Gammie, 335 F.3d 889, 900 (9th Cir.2003) (en banc). I am therefore convinced that Philips is no longer controlling.5
An equal protection analysis applying strict scrutiny to the “Don’t Ask, Don’t Tell” statute is accordingly open to us. There are two different approaches to strict scrutiny under equal protection analysis, and both should be followed in this case.
The most direct path to strict scrutiny of the statute under the equal protection principle is to hold that classifications discriminating against homosexuals are “suspect,” like classifications based on race. See Loving v. Virginia, 388 U.S. 1, 11, 87 S.Ct. 1817, 18 L.Ed.2d 1010 (1967) (subjecting race-based miscegenation statute to strict scrutiny under the Equal Protection Clause). I have long been convinced that classifications against homosexuals are suspect in the equal protection sense, but I was unable to persuade a majority of my colleagues to embark on en banc review to establish that proposition. See High Tech Gays v. Defense Industrial Security Clearance Office, 909 F.2d at 376-80 (1990) (Canby, J., dissenting from denial of rehearing en banc). As I have already explained, however, the overruling of Bowers by Lawrence has undermined High Tech Gays. We accordingly are free to revisit the question whether the adverse classification of homosexuals is “suspect” under equal protection analysis. My reasons for concluding that such classifications are suspect are fully set out in my dissent from denial of en banc review in High Tech Gays, and I will not belabor the matter here. Suffice it to say that homosexuals have “experienced a history of purposeful unequal treatment [and] been sub*825jected to unique disabilities on the basis of stereotyped characteristics not truly indicative of their abilities.” Mass. Bd. of Ret. v. Murgia, 427 U.S. 307, 313, 96 S.Ct. 2562, 49 L.Ed.2d 520 (1976) (internal quotation marks omitted). They also “exhibit obvious, immutable, or distinguishing characteristics that define them as a discrete group; and they are [ ] a minority.” Lyng v. Castillo, 477 U.S. 635, 638, 106 S.Ct. 2727, 91 L.Ed.2d 527 (1986). In short, they are a group deserving of protection against the prejudices and power of an often-antagonistic majority.
The Supreme Court’s decision in Romer v. Evans, 517 U.S. 620, 116 S.Ct. 1620, 134 L.Ed.2d 855 (1996), is not a barrier to a suspect classification, strict scrutiny approach. In that case, the Court struck down a Colorado constitutional provision prohibiting, among other things, any anti-discrimination legislation protecting homosexuals. Id. at 623-24, 116 S.Ct. 1620. The Supreme Court noted that most laws involve a classification and that, if no fundamental right or suspect class is involved, statutes are subject only to rational basis review. Id. at 631, 116 S.Ct. 1620. The Court then stated that the Colorado provision
fails, indeed defies, even this conventional inquiry. First, the amendment has the peculiar property of imposing a broad and undifferentiated disability on a single named group, an exceptional and, as we shall explain, invalid form of legislation.
Id. at 632, 116 S.Ct. 1620 (emphasis added). Thus the Court had no need to address whether homosexuals constituted a suspect class because the Colorado provision failed “even” rational basis review. That ruling does not negate the application of higher levels of scrutiny on similar classifications. Indeed, the strong language of Romer suggests that the invidiousness of the legislation would have supported any standard of review as a path to its invalidation. Romer, like Lawrence, does not forbid the application of strict scrutiny, even though it may have found that level of scrutiny unnecessary to invalidate the legislation before the Court in that case.
In addition to the avenue of a suspect classification, there is another path to strict scrutiny under equal protection analysis. Classifications that impinge on a fundamental right are subject to strict scrutiny when challenged as a violation of equal protection. See, e.g., Dunn v. Blumstein, 405 U.S. 330, 337-39, 92 S.Ct. 995, 31 L.Ed.2d 274 (1972). As I have already explained, Lawrence effectively establishes a fundamental right without so labeling it. At the very least, Lawrence leaves the question open, to permit us to recognize the fundamental right to homosexual relations as I have already insisted we must. Even though that right justifies strict scrutiny under a theory of substantive due process, there are good reasons for adding an equal protection analysis in this case. It is true that, in Lawrence, the Supreme Court elected not to employ an equal protection theory. 539 U.S. at 574-75, 123 S.Ct. 2472. It recognized, however, that equal protection provided a “tenable” basis for declaring the statute invalid, and conceded that a decision recognizing a liberty interest in certain conduct advanced the cause of equality as well as due process. Id. at 575, 123 S.Ct. 2472. The reason why the Court in Laivrence did not employ an equal protection analysis was itself protective. The Court stated that it would not sufficiently establish the right to intimate homosexual relations if only equal protection were invoked, because a state might frustrate the right by denying heterosexuals as well as homosexuals the right to non-marital sexual relations. See id.
*826The danger of an end-run remedy of equal treatment is not severe in our case, however. I doubt that the armed services are likely to respond to an invalidation of the “Don’t Ask, Don’t Tell” statute as a violation of equal protection by decreeing the automatic discharge of any member, heterosexual or homosexual, who is found to have engaged in sexual relations outside of marriage. In any event, we can guard against any such result by retaining our substantive due process analysis along with an equal protection approach.
The reason for including an equal protection analysis is that there is a very clear element of discrimination in the whole “Don’t Ask, Don’t Tell” apparatus, and an equal protection analysis focuses the inquiry sharply on a question that should not be ignored: what compelling interest of the Air Force is narrowly served by discharging homosexuals but not others who engage in sexual relations privately off duty, off base, and with persons unconnected to the military? It is no answer to such a question that the known presence of a sexually active homosexual in a military unit necessarily creates sexual tensions (if indeed that could be shown), unless it were also demonstrated that the presence of heterosexuals in a military unit created no comparable tensions. It is also not a sufficient answer that many military personnel are biased against homosexuals. See Pruitt v. Cheney, 963 F.2d 1160, 1165 (9th Cir.1992); see also Palmore v. Sidoti, 466 U.S. 429, 433, 104 S.Ct. 1879, 80 L.Ed.2d 421 (1984) (“The Constitution cannot control such prejudices but neither can it tolerate them. Private biases may be outside the reach of the law, but the law cannot, directly or indirectly, give them effect.”); Romer, 517 U.S. at 634-35, 116 S.Ct. 1620. There are other requirements of narrow tailoring that would apply during further proceedings applying strict scrutiny, but the point now is that part of the inquiry should address the clear discrimination between homosexuals and heterosexuals, and determine whether that discrimination is necessary to serve a compelling governmental interest and. sweeps no more broadly than necessary.
Order of Inquiry in Further Proceedings
The inquiry on remand should focus first on the Air Force’s justification for its impingement on the right to private intimate sexual relations and the compelling nature of any interest that is served by that measure. The Air Force should be required to identify a compelling interest with sufficient specificity so that the relation between the “Don’t Ask, Don’t Tell” statute and that policy can be evaluated. It is difficult to accomplish that goal if the compelling interest is as broadly stated as “management of the military” or, say, “winning wars.” Moreover, under strict scrutiny, it is not enough that the interest be merely “served” by the challenged legislation; the legislation must be necessary to that purpose, and must sweep no more broadly than is essential to serve the governmental purpose. See Dunn, 405 U.S at 345-46, 351-52, 92 S.Ct. 995.
Thus, as a matter of due process, the Air Force can be required to show why there is a compelling need to discharge homosexuals who have been sexually active outside of their duty station with persons unconnected to the military and why the measure it has adopted is narrowly tailored to the satisfaction of that compelling need. As a matter of equal protection, the Air Force can be asked to show what compelling need is narrowly served by treating homosexuals who are sexually active off duty and outside the military context differently from heterosexuals who are sexually active off duty and outside the military context. These requirements are case-*827specific in that they reflect the alleged facts that Major Witt conducted all of her relations with her female partner off-base, and her partner was alleged not to be in or employed by the military. If the Air Force cannot meet these requirements, the statute must be invalidated in such applications.
There are clear advantages to addressing the Air Force’s justifications first, before any inquiry into the personal characteristics and situation of Major Witt in her unit. First, requiring the Air Force to make the requisite showing as a threshold matter may end the case.
Second, the inquiry directed toward the Air Force is less potentially disruptive than a focus on Major Witt herself and, particularly, the allegedly favorable attitude toward her on the part of other members of her unit. To require unit members to testify or submit affidavits concerning the degree to which they do or do not consider themselves adversely affected by the presence of a known, sexually active homosexual, may constitute a distraction from regular duties. It is better to employ such an inquiry only as a last resort.6 Finally, requiring the Air Force to justify the application of the statute to a generic service member who carries on a homosexual relationship and intimate conduct away from the duty station and its personnel provides more protection of the constitutional right set forth in Lawrence. Because the right to choose to engage in private, intimate sexual conduct is a eonsti-tutional right of a high order, it must be protected not just for the outstanding service member like Major Witt, but also for the run-of-the-mill airman or soldier. It is thus the general application of the statute to the generic service member that the Air Force must be required to justify. In Lawrence, after all, the Supreme Court struck down the statute as applied to anyone engaging in homosexual conduct; it did not find it necessary or relevant to inquire into whether the individual conduct of which the petitioners had been convicted was more or less offensive to the interests of the State under the circumstances of its occurrence.
Conclusion
The majority opinion represents a conscientious effort to reach a just result in this case, and I agree with much of its analysis. I conclude, however, that the Air Force must demonstrate that the “Don’t Ask, Don’t Tell” statute meets the requirements of strict scrutiny — that it is necessary to serve a compelling governmental interest and that it sweeps no more broadly than necessary. I also conclude that the Air Force must be required to do so for purposes of both substantive due process and equal protection. I therefore respectfully dissent in part from the majority opinion.
. Under the facts alleged in the complaint, the statute’s popular name appears to be a misnomer as applied to Major Witt. She did not tell, but the Air Force asked.
. In that regard, Lawrence is to be contrasted with cases of gender discrimination, where the Supreme Court has expressly specified an intermediate standard of review. See Craig v. Boren, 429 U.S. 190, 197, 97 S.Ct. 451, 50 L.Ed.2d 397 (1976).
. For reasons explained in the following section on equal protection, I do not regard our earlier precedents applying lesser standards of scrutiny to military discrimination against homosexuals as binding after Lawrence. See, e.g., Beller v. Middendorf, 632 F.2d 788, 812 (9th Cir.1980) (upholding Navy policy of discharging homosexuals even though regulation is "perhaps broader than necessary”); Holmes v. Cal. Army Nat'l Guard, 124 F.3d 1126, 1132-36 (9th Cir.1997) (upholding "Don’t Ask, Don’t Tell” policy under equal protection rational basis review); Philips v. Perry, 106 F.3d 1420, 1425-29 (9th Cir.1997) (same).
. Major Witt does urge upon us a different kind of equal protection claim. She contends that the Air Force violates equal protection because it requires automatic discharge of sexually active homosexuals on the ground that they are offensive to some members of a military unit, while others equally offensive, such as child molesters, are not categorically subject to discharge. See AFI 36-3209, ¶ 2.29.10. Like the majority, I find it unnecessary to address this argument. I also conclude that it would accomplish too little to establish that persons availing themselves of their constitutional right to intimate homosexual relations should be treated at least as well as child molesters.
. In Holmes v. Cal. Army Nat'l Guard, 124 F.3d 1126, 1132 (9th Cir.1997), we also applied rational basis review to reject an equal protection challenge to a component of the “Don't Ask, Don't Tell” policy, relying on Philips and High Tech Gays. For the reasons just discussed, Lawrence's overruling of Bowers undermines Holmes as well.
. For this reason, even if I were to accept the majority's standard of scrutiny, I would modify its remand instructions now directed to determining whether "the application of DADT specifically to Major Witt significantly furthers the government’s interest....” Supra p. 821. Further proceedings should begin by requiring the Air Force to show what important governmental interest is significantly furthered by the statute. The only facts concerning Major Witt that need to be developed at that point are that her homosexual relationship was carried on off-duty, away from military premises, with a person unconnected to the military. The Air Force must then demonstrate why it is necessary to apply the statute to a service member in those circumstances. Further details of Major Witt’s individual circumstances would best be left to the end, and may be unnecessary.