American Civil Liberties Union v. Mukasey

OPINION OF THE COURT GREENBERG, Circuit Judge. I. INTRODUCTION This matter comes on before this Court on an appeal from an order of the District Court entered March 22, 2007, finding that the Child Online Protection Act (“COPA”), 47 U.S.C. § 231, facially violates the First and Fifth Amendments of the Constitution and permanently enjoining the Attorney General from enforcing COPA. The Government challenges the District Court’s conclusions that: (1) COPA is not narrowly tailored to advance the Government’s compelling interest in protecting children from harmful material on the World' Wide Web (“Web”); (2) there are less restrictive, equally effective alternatives to COPA; and (3) COPA is impermissibly overbroad and vague. We will affirm. II. FACTS AND PROCEDURAL HISTORY It is useful at the outset to set forth a short history of the background of COPA and an explanation of the relationship between the Web and the Internet. Congress enacted COPA to protect minors from exposure to sexually explicit material on the Web. The Web is just one portion of the Internet, which “is an interactive medium based on a decentralized network of computers.” American Civil Liberties Union v. Gonzales, 478 F.Supp.2d 775, 781 (E.D.Pa.2007) (“Gonzales”). “The Internet may also be used to engage in other activities such as sending and receiving emails, trading files, exchanging instant messages, chatting online, streaming audio and video, and making voice calls.” Id. The District Court described how the Web functions: On the Web, a client program called a Web browser retrieves information from the Internet, such as Web pages and other computer files using their network addresses and displays them, typically on a computer monitor.... Web pages, which can contain, inter alia, text, still and moving picture files, sound files, and computer scripts, are often arranged in collections of related material called Web sites, which consist of one or more Web pages.... It is estimated that there are between 25 and 64 billion Web pages on the surface portion of the Web (‘Surface Web’) — that is, the portion of the Web that is capable of being indexed by search engines. These Web pages may be displayed on a monitor screen and, thus, the content may be seen by anyone operating a computer or other Internet capable device which is properly connected to the Internet. Id. at 781-82 (citations omitted). The District Court indicated that “[a] little more than 1 percent of all Web pages on the Surface Web (amounting to approximately 275 million to 700 million Web pages) are sexually explicit.” Id. at 788. COPA provides for civil and criminal penalties — including up to six months imprisonment — for anyone who knowingly *185 posts “material that is harmful to minors” on the Web “for commercial purposes.” 47 U.S.C. § 231(a)(1). “Intentional” violations result in heavier fines. Id. at § 231(a)(2). “[M]aterial that is harmful to minors” includes any communication that is obscene or that: (A) the average person, applying contemporary community standards, would find, taking the material as a whole and with respect to minors, is designed to appeal to, or is designed to pander to, the prurient interest; (B) depicts, describes, or represents, in a manner patently offensive with respect to minors, an actual or simulated sexual act or sexual contact, an actual or simulated normal or perverted sexual act, or a lewd exhibition of the genitals or post-pubescent female breast; and (C) taken as a whole, lacks serious literary, artistic, political, or scientific value for minors. Id. at § 231(e)(6). “The term ‘minor’ means any person under 17 years of age.” Id. at § 231(e)(7). A person makes a communication “for commercial purposes” only if the person when making the communication “is engaged in the business of making such communications.” Id. at § 231(e)(2)(A). A person is “engaged in the business” when the person: devotes time, attention, or labor to such activities, as a regular course of such person’s trade or business, with the objective of earning a profit as a result of such activities .... [and] only if the person knowingly causes [or solicits] the material that is harmful to minors to be posted on the World Wide Web.... Id. at § 231(e)(2)(B). A Web publisher can assert an affirmative defense to prosecution under COPA if he or she: has restricted access by minors to material that is harmful to minors — (A) by requiring use of a credit card, debit account, adult access code, or adult personal identification number; (B) by accepting a digital certificate that verifies age; or (C) by any other reasonable measures that are feasible under available technology. Id. at § 231(c)(1). Congress enacted COPA after the Supreme Court declared Congress’s first attempt to protect minors from exposure to sexually explicit materials on the Web to be unconstitutional. See Reno v. American Civil Liberties Union, 521 U.S. 844, 117 S.Ct. 2329, 138 L.Ed.2d 874 (1997) (holding that the Communications Decency Act violated the First Amendment). The day after COPA became law on October 21, 1998, plaintiffs, consisting of speakers, content providers, and users of the Web, filed this action in the District Court seeking an injunction barring COPA’s enforcement. On February 1, 1999, the District Court preliminarily enjoined the Government from enforcing COPA pending a trial on the merits. American Civil Liberties Union v. Reno, 31 F.Supp.2d 473 (E.D.Pa.1999). In its opinion the court pointed out, among many other things, that the plaintiffs suggested that filtering and blocking technology was an “example of a more efficacious and less restrictive means to shield minors from harmful materials” than COPA but that the final determination of whether this was so “must await trial on the merits.” Id. at 497. The Government appealed but we affirmed the District Court’s order after concluding that the “community standards” language in section 231(e)(6)(A) by itself rendered COPA unconstitutionally over-broad. American Civil Liberties Union v. Reno, 217 F.3d 162, 173 (3d Cir.2000) (“ACL U I”). The Government then sought and obtained certiorari and the Supreme Court vacated our decision and remanded the case to us for further proceed *186 ings because the Court concluded that the “community standards” language did not, standing alone, make the statute unconstitutionally overbroad. Ashcroft v. American Civil Liberties Union, 535 U.S. 564, 585, 122 S.Ct. 1700, 1713, 152 L.Ed.2d 771 (2002). On the remand we ruled that, for a variety of reasons, COPA was not narrowly tailored to serve the Government’s compelling interest in preventing minors from being exposed to harmful material on the Web, was not the least restrictive means available to effect that interest, and was substantially overbroad. American Civil Liberties Union v. Ashcroft, 322 F.3d 240, 251-71 (3d Cir.2003) (‘ACLUII”). Consequently, we again affirmed the District Court’s order granting the preliminary injunction. Id. at 271. The Government again sought and obtained certiorari but this time the Supreme Court affirmed our decision though it remanded the case to the District Court for a trial on the merits. The Court contemplated that the record would be updated on the remand to reflect the then current technological developments and to account for any changes in the legal landscape. The Court further directed that the District Court determine whether Internet content filters are more effective than enforcement of the COPA restrictions or whether other possible alternatives are less restrictive and more effective than COPA to effectuate Congress’s intention. Ashcroft v. American Civil Liberties Union, 542 U.S. 656, 670-73, 124 S.Ct. 2783, 2794-95, 159 L.Ed.2d 690 (2004). After a bench trial, the District Court on March 22, 2007, issued extensive findings of fact, determined that plaintiffs have standing to maintain this action, and concluded that: COPA facially violates the First and Fifth Amendment rights of the plaintiffs because: (1) COPA is not narrowly tailored to the compelling interest of Congress; (2) defendant has failed to meet his burden of showing that COPA is the least restrictive and most effective alternative in achieving the compelling interest; and (3) COPA is impermissibly vague and overbroad. Gonzales, 478 F.Supp.2d at 821. The District Court permanently enjoined the Attorney General and his officers, agents, employees, and attorneys, and those persons in active concert or participation with him who received actual notice of its order, from enforcing or prosecuting matters premised upon COPA at any time for any conduct. Id. The Government then filed a timely appeal to this Court. III. JURISDICTION AND STANDARD OF REVIEW The District Court had jurisdiction under 28 U.S.C. § 1331 and we have jurisdiction pursuant to 28 U.S.C. § 1291. We review the constitutionality of a federal statute and related questions of statutory interpretation de novo. Abdul-Akbar v. McKelvie, 239 F.3d 307, 311 (3d Cir.2001). Although we generally review a district court’s factual findings for clear error, “[i]n the First Amendment context, reviewing courts have a duty to engage in a searching, independent factual review of the full record.” United States v. Scarfo, 263 F.3d 80, 91 (3d Cir.2001). The Supreme Court has emphasized that “an appellate court has an obligation to ‘make an independent examination of the whole record’ in order to make sure that ‘the judgment does not constitute a forbidden intrusion on the field of free expression.’ ” Bose Corp. v. Consumers Union of United States, Inc., 466 U.S. 485, 499, 104 S.Ct. 1949, 1958, 80 L.Ed.2d 502 (1984) (quoting *187 New York Times Co. v. Sullivan, 376 U.S. 254, 284-86, 84 S.Ct. 710, 728-29, 11 L.Ed.2d 686 (1964)). IV. DISCUSSION The First Amendment provides that “Congress shall make no law ... abridging the freedom of speech, or of the press.... ” U.S. Const, amend. I. COPA criminalizes a category of speech—“harmful to minors” material—that is constitutionally protected for adults. Because COPA is a content-based restriction on protected speech, it is presumptively invalid and the Government bears the burden of showing its constitutionality. Ashcroft, 542 U.S. at 660,124 S.Ct. at 2788. The Government challenges the District Court’s decision that COPA facially violated plaintiffs’ First Amendment rights because it was not narrowly tailored to further a compelling government interest, i .e., was not the least restrictive alternative to advance that interest, the prevention of minors from being exposed to harmful material on the Web, and was impermissibly vague and overbroad. 1 A. Law-of-the-Case Doctrine Before we reach the merits of the case, we must address the effect of our prior decision in ACLU II on this appeal, as the presence of that decision may make the law-of-the-case doctrine relevant here. Under the law-of-the-case doctrine, “when a court decides upon a rule of law, that decision should continue to govern the same issues in subsequent stages in the same case.” Christianson v. Colt Indus. Operating Corp., 486 U.S. 800, 816, 108 S.Ct. 2166, 2177, 100 L.Ed.2d 811 (1988) (quoting Arizona v. California, 460 U.S. 605, 618, 103 S.Ct. 1382, 1391, 75 L.Ed.2d 318 (1983)). “This rule of practice promotes the finality and efficiency of the judicial process by protecting against the agitation of settled issues.” Id. (citation and quotation marks omitted). We recently addressed the binding effect that our prior decisions on legal issues at the preliminary injunction stage on an earlier appeal in the same case have on later decisions. See Pitt News v. Pappert, 379 F.3d 96, 104-05 (3d Cir.2004). Clearly the nature of the showing that an applicant for a preliminary injunction must make to obtain relief can present special difficulties in applying the law-of-the-case doctrine in later stages of the litigation. In Pitt News we noted that “three separate rules are relevant” when considering the effect of a preliminary injunction later in ongoing litigation: First, it is our Court’s tradition that a panel may not overrule ‘a holding’ of a prior panel. Second, it is well established that neither this tradition nor the law-of-the-case doctrine requires a panel hearing an appeal from the entry of a final judgment to follow the legal analysis contained in a prior panel decision addressing the question whether a party that moved for preliminary injunctive relief showed a likelihood of success on the merits. Third, although a panel entertaining a preliminary injunction appeal generally decides only whether the district court abused its discretion in ruling on the request for relief and generally does not go into the merits any farther than is necessary to determine whether the moving party established a likelihood of success, a panel is not always required to take this narrow approach. If a preliminary injunction ap *188 peal presents a question of law and the facts are established or of no controlling relevance, the panel may decide the merits of the claim. Id. at 104-05 (citations and most internal quotation marks omitted). We explained: In the typical situation—where the prior panel stopped at the question of likelihood of success—the prior panel’s legal analysis must be carefully considered, but it is not binding on the later panel. Indeed, particularly where important First Amendment issues are raised, the later panel has a duty, in the end, to exercise its own best judgment. On the other hand, if the first panel does not stop at the question of likelihood of success and instead addresses the merits, the later panel, in accordance with our Court’s traditional practice, should regard itself as bound by the prior panel opinion. Id. at 105. But even if we subsequently conclude that in a particular case our prior determination ordinarily would bind us, we may reconsider issues that we previously resolved if any of the following “extraordinary circumstances” are present: “(1) there has been an intervening change in the law; (2) new evidence has become available; or (3) reconsideration is necessary to prevent clear error or a manifest injustice.” Council of Alternative Political Parties v. Hooks, 179 F.3d 64, 69 (3d Cir.1999) (citing In re City of Philadelphia Litig., 158 F.3d 711, 718 (3d Cir.1998)). In ACLU II we concluded that plaintiffs were likely to succeed on the merits and thus concluded that the District Court could grant them a preliminary injunction. Nevertheless we did not stop our analysis after coming to that conclusion. Instead, we opined at length on the constitutionality of COPA and construed a number of terms of the statute. Consequently, the procedural posture of this case and the scope of our prior decision has set a foundation for the possible applicability of the law-of-the-case doctrine here. Though we will explain in more detail the basis for our conclusions in ACLU II, for purposes of determining the binding effect of that decision on this appeal it is enough to note now that we expressly held the following: (1) COPA’s definitions of “material that is harmful to minors,” and “commercial purposes” and COPA’s affirmative defenses are not narrowly tailored to achieve the Government’s compelling interest in protecting minors from harmful material on the Web, 322 F.3d at 251; (2) filtering software is a less restrictive alternative than the COPA restrictions to advance the Government’s compelling interest in preventing minors from being exposed to harmful material on the Web, id. at 265; (3) COPA is “substantially overbroad” because of its use of the terms “material harmful to minors,” “minor,” “commercial purposes,” and “community standards”; (4) COPA’s affirmative defenses do not save the statute from sweeping too broadly; and (5) a narrowing construction of COPA is not available to permit it to be upheld, id. at 266-71. In its decision affirming ACLU II, the Supreme Court expressly declined to consider many of the issues that we had determined. Specifically, the Court stated: [W]e agree with the Court of Appeals that the District Court did not abuse its discretion in entering the preliminary injunction. Our reasoning in support of this conclusion, however, is based on narrower, more specific grounds than the rationale the Court of Appeals adopted. The Court of Appeals, in its opinion affirming the decision of the District Court, construed a number of *189 terms in the statute, and held that COPA, so construed, was unconstitutional. None of those constructions of statutory terminology, however, were relied on by or necessary to the conclusions of the District Court. Instead, the District Court concluded only that the statute was likely to burden some speech that is protected for adults, which [the Government] does not dispute. As to the definitional disputes, the District Court concluded only that [the plaintiffs’] interpretation was ‘not unreasonable,’ and relied on their interpretation only to conclude that [the plaintiffs] had standing to challenge the statute, which, again, [the Government] does not dispute. Because we affirm the District Court’s decision to grant the preliminary injunction for the reasons relied on by the District Court, we decline to consider the correctness of the other arguments relied on by the Court of Appeals. Ashcroft, 542 U.S. at 665, 124 S.Ct. at 2791 (citations omitted). The Court then addressed the issue of whether there are less restrictive alternatives to the COPA restrictions to further the Government’s compelling interest in COPA’s objective and stated that “[f]ilters are less restrictive than COPA.” Id. at 667, 124 S.Ct. at 2792. The Court recognized, however, that “there are substantial factual disputes remaining in the case.... [T]here is a serious gap in the evidence as to the effectiveness of filtering software. For us to assume, without proof, that filters are less effective than COPA would usurp the District Court’s factfinding role.” Id. at 671, 124 S.Ct. at 2794 (citation omitted). Thus, the Court recognized that restrictiveness and effectiveness are separate matters. The Court also noted that: [T]he factual record does not reflect current technological reality — a serious flaw in any case involving the Internet. The technology of the Internet evolves at a rapid pace. Yet the factfindings of the District Court were entered in February 1999, over five years ago.... It is reasonable to assume that other technological developments important to the First Amendment analysis have also occurred during that time. More and better filtering alternatives may exist than when the District Court entered its findings. Id. Accordingly, the Court decided to remand the case to the District Court for a full trial on the merits to “update and supplement the factual record to reflect current technological realities” and “to take account of a changed legal landscape” to determine if other methods were less restrictive alternatives to COPA to further the Government’s compelling interest in its objective. Id. at 672, 124 S.Ct. at 2795. The Government contends that the portion of our opinion in ACLU II that goes beyond the Supreme Court’s holding “is not binding because the Supreme Court’s decision remanding for further consideration of the question whether filtering is a less restrictive alternative than COPA contemplates a fresh examination of all the issues in this case, including the scope of COPA’s coverage and its efficacy and restrictiveness compared to filtering.” Appellant’s Letter at 1 (May 30, 2008). 2 We conclude, however, that the Government is incorrect on this point. The Supreme Court’s decision explicitly left untouched our conclusions in ACLU II other than our decision that filters are a less restrictive alternative than COPA for advancing the Government’s compelling interest at stake in this litigation. Moreover, our other determinations — including our interpretation *190 of the provisions of COPA and whether they are narrowly construed or impermis-sibly overbroad—did not depend' on the factual record and thus would not be implicated by the evidence developed in the subsequent trial on the merits in the District Court. Accordingly, those conclusions remain binding on us now. The Government also contends that we should reconsider the issues addressed in ACLU II on the basis of an intervening change in the law since we decided that case. In this regard it points to the Supreme Court’s recent decision in United States v. Williams, — U.S. -, 128 S.Ct. 1830, 170 L.Ed.2d 650 (2008), where the Court found that the Prosecutorial Remedies and Other Tools to end the Exploitation of Children Today Act of 2003, 18 U.S.C. § 2252A(a)(3)(B), is not over-broad under the First Amendment. But the Court in Williams merely restated and applied the well-established legal doctrines of overbreadth and vagueness and did not change the law applicable to this case. Accordingly, we conclude that there are not “extraordinary circumstances” justifying us in departing from our holdings in ACLU II other than that with respect to filtering. Now that we have delineated the contours of ACLU II’s effect on this appeal, we will address the issues the Government raises. As we consider these issues, we will determine whether, and if so the extent, that our conclusions in ACLU II are the law-of-the-case here. B. Strict Scrutiny First, the Government challenges the District Court’s decision that COPA is unconstitutional because it does not survive strict scrutiny, the standard that we apply in this case inasmuch as COPA is a content-based restriction on speech. See Turner Broadcasting Sys., Inc. v. Fed. Commc’ns Comm’n, 512 U.S. 622, 642, 114 S.Ct. 2445, 2459, 129 L.Ed.2d 497 (1994). To survive strict scrutiny analysis, a statute must: (1) serve a compelling governmental interest; (2) be narrowly tailored to achieve that interest; and (3) be the least restrictive means of advancing that interest. Sable Commc’ns of Cal., Inc. v. Fed. Commc’ns Comm’n, 492 U.S. 115, 126, 109 S.Ct. 2829, 2836, 106 L.Ed.2d 93 (1989). 1. Compelling Interest As we noted above, Congress enacted COPA to protect minors from exposure to sexually explicit material on the Web. The Supreme Court has held that “there is a compelling interest in protecting the physical and psychological well-being of minors,” Sable, 492 U.S. at 126, 109 S.Ct. at 2836, and the parties agree that the Government has a compelling interest to protect minors from exposure to harmful material on the Web. Inasmuch as we agree with them on that point, we turn to the question of whether COPA is narrowly tailored to effectuate its purpose. 2. Narrowly Tailored As we stated above, to survive a strict scrutiny analysis COPA must be narrowly tailored to advance a compelling government interest. In ACLU II, we addressed this issue and held that the following provisions of COPA are not narrowly tailored: (a) the definition of ‘material that is harmful to minors,’ which includes the concept of taking ‘as a whole ’ material designed to appeal to the ‘prurient interest’ of minors; and material which (when judged as a whole) lacks ‘serious literary’ or other ‘value’ for minors', (b) the definition of ‘commercial purposes,’ which limits the reach of the statute to persons ‘engaged in the business ’ *191 (broadly defined) of making communications of material that is harmful to minors; and (c) the ‘affirmative defenses’ available to publishers, which require the technological screening of users for the purpose of age verification. ACLU II, 322 F.3d at 251. First, we addressed why we found that the “taking the material as a whole” language in COPA’s definition of “material that is harmful to minors,” was not narrowly tailored. COPA defines such material to include any matter that is obscene or that: (A) the average person, applying contemporary community standards, would find, taking the material as a whole and with respect to minors, is designed to appeal to, or is designed to pander to, the prurient interest; (B) depicts, describes, or represents, in a manner patently offensive with respect to minors, an actual or simulated sexual act or sexual contact, an actual or simulated normal or perverted sexual act, or a lewd exhibition of the genitals or post-pubescent female breast; and (C) taken as a whole, lacks serious literary, artistic, political, or scientific value for minors. 47 U.S.C. § 231(e)(6) (emphasis added). We concluded that the taken “as a whole” language, when read in context with other language in the statute, mandates evaluation of an exhibit on the Internet in isolation, rather than in context. ACLU II, 322 F.3d at 253. We explained that: Because we view such a statute, construed as its own text unquestionably requires, as pertaining only to single individual exhibits, COPA endangers a wide range of communications, exhibits, and speakers whose messages do not comport with the type of harmful materials legitimately targeted under COPA, i.e., material that is obscene as to minors. Accordingly, while COPA penalizes publishers for making available improper material for minors, at the same time it impermissibly burdens a wide range of speech and exhibits otherwise protected for adults. Thus, in our opinion, the Act, which proscribes publication of material harmful to minors, is not narrowly tailored to serve the Government’s stated purpose in protecting minors from such material. Id. (citation omitted). We also explained why we found that “COPA’s definition of the term ‘minor,’ viewed in conjunction with the ‘material harmful to minors’ test, is not tailored narrowly enough to satisfy the First Amendment’s requirements.” Id. at 255. COPA defines “minor” as “any person under 17 years of age.” 47 U.S.C. § 231(e)(7). We stated that the term “thus applies in a literal sense to an infant, a five-year old, or a person just shy of age seventeen.” ACLU II, 322 F.3d at 254. We reasoned that ‘Web publishers would face great uncertainty in deciding what minor could be exposed to its publication, so that a publisher could predict, and guard against, potential liability.” Id. at 255. We explicitly rejected the Government’s argument that the term “should be read to apply only to normal, older adolescents,” id. at 254, and stated that under either our definition or the Government’s proffered definition, “the term ‘minor,’ viewed in conjunction with the ‘material harmful to minors’ test, is not tailored narrowly enough to satisfy the First Amendment’s requirements,” id. at 255. We then proceeded to explain why we found that “COPA’s purported limitation of liability to persons making communications ‘for commercial purposes’ does not narrow the reach of COPA sufficiently.” Id. at 256. COPA states that “[a] person shall be considered to make a communication for commercial purposes only if such *192 person is engaged in the business of making such communications,” and that [t]he term ‘engaged in the business’ means that the person who makes a communication ... that includes any material that is harmful to minors, devotes time, attention, or labor to such activities, as a regular course of such person’s trade or business, with the objective of earning a profit.... A person may be considered to be engaged in the business ... only if the person knowingly causes [or solicits] the material that is harmful to minors to be posted on the World Wide Web .... 47 U.S.C. § 231(e)(2). We stated that: we read COPA to apply to Web publishers who have posted any material that is ‘harmful to minors’ on their Web sites, even if they do not make a profit from such material itself or do not post such material as the principal part of their business. Under the plain language of COPA, a Web publisher will be subjected to liability if even a small part of his or her Web site displays material ‘harmful to minors.’ ACLU II, 322 F.3d at 256. We stated that this group included “those persons who sell advertising space on their otherwise noncommercial Web sites ... [, including] the Web publisher who provides free content on his or her Web site and seeks advertising revenue, perhaps only to defray the cost of maintaining the Web site.” Id. We also rejected the Government’s argument that “COPA’s definition of ‘engaged in the business’ limits liability to those persons who publish material that is harmful to minors ‘as a regular course of such person’s business or trade’ COPA’s use of the phrase ‘regular course’ does not narrow the scope of speech covered because it does not place any limitations on the amount, or the proportion, of a Web publisher’s posted content that constitutes such material. Thus, even if posted material that is harmful to minors constitutes only a very small, or even infinitesimal, part of a publisher’s entire Web site, the publisher may still be subject to liability. Id. at 257. Finally, we explained why we found that COPA’s affirmative defenses were not narrowly tailored. As we already have noted above, a Web publisher can assert an affirmative defense if it: has restricted access by minors to material that is harmful to minors — (A) by requiring use of a credit card, debit account, adult access code, or adult personal identification number; (B) by accepting a digital certificate that verifies age; or (C) by any other reasonable measures that are feasible under available technology. 47 U.S.C. § 231(c)(1). We first stated that implementation of the affirmative defenses in COPA “will likely deter many adults from accessing restricted content, because many Web users are simply unwilling to provide identification information in order to gain access to content, especially where the information they wish to access is sensitive or controversial.” ACLU II, 322 F.3d at 259 (footnote omitted). For this particular conclusion we relied on factual findings the District Court made in granting the preliminary injunction, so to this extent it does not bind us on this appeal. Though we are not bound by previous conclusions with respect to deterrence of adults seeking restricted content, in ACLU II we reached other conclusions about COPA’s affirmative defenses that do not depend on the facts as developed in the District Court, and those conclusions are binding on us on this appeal. For instance, in ACLU II we stated that “the affirmative defenses do not provide Web publishers with assurances of freedom *193 from prosecution” because “ ‘[a]n affirmative defense applies only after prosecution has begun, and the speaker must himself prove ... that his conduct falls within the affirmative defense.’” Id. at 260 (second alteration in original) (quoting Ashcroft v. Free Speech Coalition, 535 U.S. 234, 255, 122 S.Ct. 1389, 1404, 152 L.Ed.2d 403 (2002)). We also considered the Government’s argument that other cases dealing with display restrictions have upheld the use of blinder racks to shield minors from viewing harmful material. We distinguished those cases because: [t]he use of ‘blinder racks’ ... does not create the same deterrent effect on adults as would COPA’s credit card or adult verification screens. Blinder racks do not require adults to compromise their anonymity in their viewing of material harmful to minors, nor do they create any financial burden on the user. Moreover, they do not burden the speech contained in the targeted publications any more than is absolutely necessary to shield minors from its content. Id. We concluded that “[t]he effect of the affirmative defenses, as they burden ‘material harmful to minors’ which is constitutionally protected for adults, is to drive this protected speech from the marketplace of ideas on the Internet. This type of regulation is prohibited under the First Amendment.” Id. In its decision made after the trial on the merits now on appeal before us, the District Court concluded that COPA is not narrowly tailored because it is both overin-clusive and underinclusive. First, the court determined that COPA is impermis-sibly overinclusive because it “prohibits much more speech than is necessary to further Congress’ compelling interest. For example, ... the definitions of ‘commercial purposes’ and ‘engaged in the business’ apply to an inordinate amount of Internet speech and certainly cover more than just commercial pornographers.... ” Gonzales, 478 F.Supp.2d at 810 (citations omitted). The court also concluded that COPA is overinclusive because it “applies to speech that is obscene as to all minors from newborns to age sixteen, and not just to speech that is obscene as to older minors ....” Id. The Government contends that COPA is narrowly tailored because it applies only to commercial pornographers and only to material that is harmful to “older” minors. But we addressed and rejected the Government’s arguments in ACLU II, when we found there is nothing in the text of COPA to limit its application solely to “commercial pornographers” or to limit the phrase “material that is harmful to minors” to include material that only is harmful to “older” minors. See 322 F.3d at 253-57. Our prior decision is binding on these issues on this appeal. The District Court also found that COPA is not narrowly tailored because it is underinclusive. In ACLU II we did not address whether COPA is impermissibly underinclusive and so we are free to review this finding on the merits. In its Findings of Fact, the District Court stated that “a substantial number (approximately 50 percent) of sexually explicit websites are foreign in origin.” Gonzales, 478 F.Supp.2d at 789. The court then reasoned: [TJhere is a significant amount of sexually explicit material on the Internet which originates from outside of the United States.... [Ujnlike Internet content filters which are able to block from view unsuitable material regardless of its origin, COPA has no extraterritorial application. As a result, ... COPA is not applicable to a large amount of material that is unsuitable for children which originates overseas but is *194 nevertheless available to children in the United States_COPA’s lack of extraterritorial application renders it under-inclusive. Id. at 810-11 (citations omitted). The Government contends that the District Court erred by construing COPA not to apply to foreign Web sites, and thus the Government argues that COPA is not un-derinclusive. The problem with the Government’s argument in this respect is that, as we explain below, the Supreme Court already has determined that COPA does not apply to foreign Web sites. But notwithstanding this significant limitation on COPA’s scope, if we had to pass on the issue we might conclude that COPA is not unconstitutionally underinclusive. The Supreme Court has explained the circumstances in which a court may find that a regulation of speech is impermissibly underinclusive: [A]n exemption from an otherwise permissible regulation of speech may represent a governmental ‘attempt to give one side of a debatable public question an advantage in expressing its views to the people.’ First Nat’l Bank of Boston v. Bellotti 435 U.S. 765, 785-86, 98 S.Ct. 1407, 1420-21, 55 L.Ed.2d 707 (1978): Alternatively, through the combined operation of a general speech restriction and its exemptions, the government might seek to select the ‘permissible subjects for public debate’ and thereby to ‘control ... the search for political truth.’ Consolidated Edison Co. of N.Y. v. Public Serv. Comm’n of N.Y., 447 U.S. 530, 538, 100 S.Ct. 2326, 2333, 65 L.Ed.2d 319 (1980). City of Ladue v. Gilleo, 512 U.S. 43, 51, 114 S.Ct. 2038, 2043, 129 L.Ed.2d 36 (1994) (second alteration in original) (footnote omitted). These quite narrow circumstances are hardly applicable to COPA. Even though, as the District Court recognized, COPA does not apply to foreign Web sites, we cannot understand how that limitation on its scope would “represent a governmental attempt to give one side of a debatable public question an advantage in expressing its views to the people ... [or] to select the permissible subjects for public debate.” Id. (citations and quotation marks omitted). There is no evidence in the record of which we are aware that Congress sought to favor foreign Web site publishers over domestic Web site publishers when regulating sexually explicit material on the Web, nor is there any suggestion in the record that the Government is selecting the permissible subject for public debate by excluding foreign Web sites from COPA’s coverage. In fact, we think that it is likely that Congress would have desired to place COPA’s restrictions on foreign Web sites available for access in this country but chose not to do so because, as the District Court recognized: [enforcement of COPA against overseas Web site owners would ... be burdensome and impractical due to the knotty questions of jurisdiction which arise in the Internet context. Furthermore, even if a specific foreign Web site had sufficient contacts with the forum to allow personal jurisdiction, it could be quite difficult or impossible to ensure that the offender would obey or could be forced to obey the judgment of the U.S. court. Gonzales, 478 F.Supp.2d at 811. In these circumstances, even though COPA’s omission of foreign Web sites from its regulations certainly is relevant in an inquiry into whether it is the most effective means of advancing the Government’s compelling interest in COPA’s object, the omission might not lead us to a conclusion that the statute is impermissibly underinclusive. After all, as the Court of Appeals for the *195 Eighth Circuit recently noted, “a limitation on speech that is not all-encompassing may still be narrowly tailored where the under-inclusivity does not favor a particular viewpoint or undermine the rationale given for the regulation.” Bowman v. White, 444 F.3d 967, 983 (8th Cir.2006). On the other hand, we might conclude that because COPA fails to apply to 50% of its purported commercial pornography targets, we lack the evidence necessary to satisfy us that Congress had in mind its stated goal of protecting minors from harmful material on the Web when it passed COPA. It is not as though Congress is unable to protect minors from harmful material on foreign Web sites; for instance, Congress could promote the use of Internet content filters, which do not discriminate on the basis of geography. COPA’s failure to protect minors from harmful material on foreign Web sites might raise the inference that Congress had some ulterior, impermissible motive for passing COPA. We note, however, that our possible disagreement with the District Court on this one point would not change our ultimate decision to affirm its order granting a permanent injunction, as there are numerous other grounds that require us to find that COPA is not narrowly tailored and is unconstitutional. Accordingly, we will refrain from deciding the matter. The District Court also found that COPA’s affirmative defenses “do not aid in narrowly tailoring COPA to Congress’ compelling interest.” Gonzales, 478 F.Supp.2d at 813. Specifically, the court found that: there is no evidence of age verification services or products available on the market to owners of Web sites that actually reliably establish or verify the age of Internet users. Nor is there evidence of such services or products that can effectively prevent access to Web pages by a minor. Id. at 800. The court found that “[t]he rules of payment card associations in this country prohibit Web sites from claiming that use of a payment card is an effective method of verifying age, and prohibit Web site owners from using credit or debit cards to verify age,” and that “a significant number of minors have access to [payment cards].” Id. at 801. The court also reviewed data verification services, which are “non-payment card-based services that attempt to verify the age or identity of an individual Internet user,” and found that they are unreliable because they “cannot determine whether the person entering information into the Web site is the person to whom the information pertains.” Id. at 802. The court further found that the minimum information required by a data verification services company “can easily be circumvented by children who generally know the first and last name, street address and zip codes of their parents or another adult.” Id. The court later explained, “[t]he affirmative defenses cannot cure COPA’s failure to be narrowly tailored because they are effectively unavailable. Credit cards, debit accounts, adult access codes, and adult personal identification numbers do not in fact verify age. As a result, their use does not, in good faith, ‘restrict [ ] access’ by minors.” Id. at 811 (second alteration in original) (quoting 47 U.S.C. § 231(c)(1)(A)). The court also concluded that COPA’s affirmative defenses “raise unique First Amendment issues” that make the statute unconstitutional. Id. at 813. The court found that due to the fees associated with the use of the procedures enumerated in all of the affirmative defenses and verification services, “Web sites ... which desire to provide free distribution of their infor *196 mation, will be prevented from doing so.” Id. at 804. The court also found that: [f]or a plethora of reasons including privacy and financial concerns ... and the fact that so much Web content is available for free, many Web users already refuse to register, provide credit card information, or provide real personal information to Web sites if they have any alternative. Because requiring age verification would lead to a significant loss of users, content providers would have to either self-censor, risk prosecution, or shoulder the large financial burden of age verification. Id. at 805. Moreover, the court found that “many users who are not willing to access information non-anonymously will be deterred from accessing the desired information. Web site owners ... will be deprived of the ability to provide this information to those users.” Id. at 806. The court also indicated that: [Requiring Internet users to provide payment card information or other personally identifiable information to access a Web site would significantly deter many users from entering the site, because Internet users are concerned about security on the Internet and because Internet users are afraid of fraud and identity theft on the Internet. Id. Based on these findings, the court concluded that: [t]he affirmative defenses also raise their own First Amendment concerns. For example, the utilization of those devices to trigger COPA’s affirmative defenses will deter listeners, many of whom will be unwilling to reveal personal and financial information in order to access content and, thus, will chill speech. Similarly, the affirmative defenses also impermissibly burden Web site operators with demonstrating that their speech is lawful. Under the COPA regime, Web site operators are unable to defend themselves until after they are prosecuted. Moreover, the affirmative defenses place substantial economic burdens on the exercise of protected speech because all of them involve significant cost and the loss of Web site visitors, especially to those plaintiffs who provide their content for free. Id. at 812-13 (citations and quotations omitted). The Government argues that the District Court erred in rejecting the limiting effect of COPA’s affirmative defenses. It contends that “[t]he possibility that some minors may have access to credit cards merely demonstrates that no system of age verification is foolproof. It does not call into question the availability of credit card screening as an affirmative defense that tailors COPA more narrowly.” Appellant’s Br. at 37. The Government also argues that “the court ignored testimony that minors do not have access to traditional payment cards under their own control but simply have access to cards supervised by adults.” Id. But the District Court found that even if there is parental supervision of payment card use, the supervision does not prevent access to harmful material by minors because parents “may not be able to identify transactions on sexually explicit Web sites because the adult nature of such transactions is often not readily identifiable.... ” Gonzales, 478 F.Supp.2d at 802. In any event, we conclude that the District Court correctly found that the affirmative defenses are “effectively unavailable” because they do not actually verify age. The Government also argues that the District Court incorrectly determined that the affirmative defenses present their own First Amendment concerns by imposing undue burdens on Web publishers due to the high costs of implementing age verifi *197 cation technologies and the loss of traffic that would result from the use of these technologies. The Government contends that the: court’s evaluation of the burdens imposed by COPA was flawed because the court focused largely, if not exclusively, on Web publishers who provide their content for free. Whatever limited application COPA might have beyond its core regulation of commercial pornography, the court erred in evaluating the burdens the statute imposes based entirely on these marginal cases and ignoring the heartland of the statute’s proscriptions, where the burdens are far less onerous. Appellant’s Br. at 38-39 (citations and quotations omitted). We reject this argument. The fact that COPA places burdens on Web publishers whom the Government does not consider to be within the “heartland” of the statute does not make those burdens any less onerous or offensive to the principles of the First Amendment. Moreover, there is good reason to believe that COPA unduly would burden even those Web publishers whom the Government considers to fall within the “heartland” of the statute, because the District Court found that those publishers also will face significant costs to implement the affirmative defenses and will suffer the loss of legitimate visitors once they do so. And, contrary to the Government’s suggestion at oral argument, users would have alternatives to obtain pornography even if COPA was in effect because, as we already have indicated and discuss below, COPA does not apply to foreign Web sites. The loss of traffic that would result clearly is an undue burden on even those Web sites that the Government contends are in the “heartland” of COPA. We conclude that the District Court correctly found that implementation of COPA’s affirmative defenses by a Web publisher so as to avoid prosecution would involve high costs and also would deter users from visiting implicated Web sites. It is clear that these burdens would chill protected speech and thus that the affirmative defenses fail a strict scrutiny analysis. The Government contends that nevertheless these burdens “are no different in kind or degree from the burdens imposed by state laws regulating the sale and commercial display of ‘harmful to minors’ materials .... [T]he effect of the statute is simply to requirfe] the commercial pornographer to put sexually explicit images behind the counter.” Appellant’s Br. at 43 (citations and certain internal quotation marks omitted) (second alteration in original). We rejected this argument in ACLU II. See 322 F.3d at 260 (“Blinder racks do not require adults to compromise their anonymity in their viewing of material harmful to minors, nor do they create any financial burden on the user. Moreover, they do not burden the speech contained in the targeted publications any more than is absolutely necessary to shield minors from its content.”). Blinder racks do not require adults to pay for speech that otherwise would be accessible for free, they do not require adults to relinquish their anonymity to access protected speech, and they do not create a potentially permanent electronic record. Blinder racks simply do not involve the privacy and security concerns that COPA’s affirmative defenses raise, and so the Government’s attempted analogy is ill-fitting. In sum, after considering our previous conclusions in ACLU II and our analyses of the issues ACLU II has not resolved, we are quite certain that notwithstanding Congress’s laudable purpose in enacting COPA, the Government has not met its *198 burden of showing that it is narrowly tailored so as to survive a strict scrutiny analysis and thereby permit us to hold it to be constitutional. 3. Least Restrictive Alternative In addition to failing the strict scrutiny test because it is not narrowly tailored, COPA does not employ the least restrictive alternative to advance the Government’s compelling interest in its purpose, the third prong of the three-prong strict scrutiny test. “A statute that ‘effectively suppresses a large amount of speech that adults have a constitutional right to receive and to address to one another ... is unacceptable if less restrictive alternatives would be at least as effective in achieving the legitimate purpose that the statute was enacted to serve.’ ” Ashcroft, 542 U.S. at 665, 124 S.Ct. at 2791 (alteration in original) (quoting Reno, 521 U.S. at 874, 117 S.Ct. at 2346). “[T]he burden is on the Government to prove that the proposed alternatives will not be as effective as the challenged statute.” Id. (citing Reno, 521 U.S. at 874, 117 S.Ct. at 2346). The Government’s burden is “not merely to show that a proposed less restrictive alternative has some flaws; its burden is to show that it is less effective.” Id. at 669, 124 S.Ct. at 2793 (citing Reno, 521 U.S. at 874,117 S.Ct. at 2346). Based on the preliminary injunction record in this case, the Supreme Court held that “[b]locking and filtering software is an alternative that is less restrictive than COPA, and, in addition, likely more effective as a means of restricting children’s access to materials harmful to them.” Id. at 666-67, 124 S.Ct. at 2792. We reached a similar conclusion in ACLU II. See 322 F.3d at 265. 3 After the trial on the merits, the District Court concluded that the Government did not meet its burden of showing that COPA is the least restrictive effective alternative for advancing Congress’s compelling interest because filter software and the Government’s pro- *199 motion and support of filter software is a less restrictive effective alternative to COPA. The District Court discussed Internet content filters at length in its Findings of Fact. We will review these findings in detail, as the need to determine whether filters are more effective than COPA to effectuate Congress’s purpose in enacting that statute was the primary reason the Supreme Court remanded the case. According to the District Court: Internet content filters (‘filters’) are computer applications which, inter alia, attempt to block certain categories of material from view that a Web browser or other Internet application is capable of displaying or downloading, including sexually explicit material. Filters categorize and block Web sites or pages based on their content. By classifying a site or page, and refusing to display it on the user’s computer screen, filters can be used to prevent children from seeing material that might be considered unsuitable. Gonzales, 478 F.Supp.2d at 789. The court explained: Filters can be programmed or configured in a variety of different ways according to, inter alia, the values of the parents using them and the age and maturity of their children.... [F]ilters can be set up to restrict materials available on Web pages and other Internet applications based on numerous factors including the type of content they contain, the presence of particular words, the address of the Web site, the Internet protocol used, or computer application used. Some filters can also restrict Internet access based on time of day, day of week, how long the computer has been connected to the Internet, or which user is logged onto a computer. Id. at 790. The court then described in detail how filters operate: Filters use different mechanisms to attempt to block access to material on the Internet including: black lists, white lists, and dynamic filtering. Black lists are lists of URLs or Internet Protocol (‘IP’) addresses that a filtering company has determined lead to content that contains the type of materials its filter is designed to block. White lists are lists of URLs or IP addresses that a filtering company has determined do not lead to any content its filter is designed to block, and, thus, should never be blocked.... In addition to its own black and white lists, filters often give parents or administrators the option of creating customized black or white lists. Dynamic filtering products use artificial intelligence to analyze Web site content in real-time as it is being requested and determine whether it should be blocked by evaluating a number of different parts of the content, both what the user can actually see on the Web page, and the various hidden pieces of information contained with the content that are part of its software code or script, known as the ‘metadata.’ Among other things, dynamic filters analyze the words on the page, the metadata, the file names for images, the URLs, the links on a page, the size of images, the formatting of the page, and other statistical pattern recognition features, such as the spatial patterns between certain words and images, which can often help filters categorize content even if the actual words are not recognized. In addition to analyzing the content of Web pages, dynamic filters also take the context of the page into consideration, to ensure that the determinations are as accurate as possible. For example, many companies will develop templates that provide additional context to teach the software how to *200 recognize certain eontexts-for example, to block the word ‘breast’ when used in combination with the word ‘sexy,’ but not when used in combination with the words ‘chicken’ or ‘cancer.’ The software analyzes context, in part, by utilizing statistical pattern recognition techniques to identify common features of acceptable and unacceptable Web pages, depending on the context in which the content appears. Id. at 790-91 (citations omitted). The court found that: [filters can be used by parents to block material that is distributed on the Web and on the other widely used parts of the Internet through protocols other than HTTP and through other Internet applications. For example, filters can be used to block any Internet application, including email, chat, instant messaging, peer-to-peer file sharing, newsgroups, streaming video and audio, Internet television and voice over Internet protocol (‘VoIP’), and other Internet protocols such as FTP. In addition to blocking access to these Internet applications completely, some products provide parents with the option of providing limited access to these applications. For example, instant messaging and email may be permitted, but some of the filtering products will only permit the sending and receiving of messages from certain authorized individuals, and will block e-mails or instant messages containing inappropriate words or any images. Filtering programs can also completely prevent children from entering or using chat rooms, or some can merely filter out any inappropriate words that come up during a chat session. Id. at 791 (citations omitted). The court then described the flexible nature of filters: Some filtering programs offer only a small number of settings, while others are highly customizable, allowing a parent to make detailed decisions about what to allow and what to block. Filtering products do this by, among other things, enabling parents to choose which categories of speech they want to be blocked (such as sexually explicit material, illicit drug information, information on violence and weapons, and hate speech) and which age setting they want the product to apply.... Filtering products can be used by parents even if they have more than one child. For example, if a family has four children, many filtering products will enable the parent to set up different accounts for each child, to ensure that each child is able to access only the content that the parents want that particular child to access. Id. (citations omitted). The court found that: [filtering products block both Web pages originating from within the United States and Web pages originating from outside the United States. The geographic origin of a Web page is not a factor in how a filter works because the filter analyzes the content of the Web page, not the location from which it came. Id. at 791-92. The court found that “[filtering products block both non-commercial and commercial Web pages.” Id. at 792. The court also found that: [i]n addition to their content filtering features, filtering products have a number of additional tools to help parents control their children’s Internet activities. Other tools available to parents include monitoring and reporting features that allow supervising adults to know which sites a minor has visited and *201 what other types of activities a minor has engaged in online. Id. The District Court found that “[filters are widely available and easy to obtain,” and that “filtering programs are fairly easy to install, configure, and use and require only minimal effort by the end user to configure and update.” Id. at 793. The court found that “installing and setting up a filter will usually take a typical computer user no more than ten or fifteen minutes. The installation and set-up process is not technically complex and does not require any special training or knowledge.” Id. at 794. The court then considered the evidence regarding the effectiveness of filters. It found that: [filtering products have improved over time and are now more effective than ever before. This is because, as with all software, the filtering companies have addressed problems with the earlier versions of the products in an attempt to make their products better. Another reason the effectiveness of filtering products has improved is that many products now provide multiple layers of filtering. Whereas many filters once only relied on black lists or white lists, many of today’s products utilize black lists, white lists, and real-time, dynamic filtering to catch any inappropriate sites that have not previously been classified by the product. There is a high level of competition in the field of Internet content filtering. That factor, along with the development of new technologies, has also caused the products to improve over time. Id. at 794-95 (citations omitted). The District Court then found that: [o]ne of the features of filtering programs that adds to their effectiveness is that they have built-in mechanisms to prevent children from bypassing or circumventing the filters, including password protection and other devices to prevent children from uninstalling the product or changing the settings. Some products even have a tamper detection feature, by which they can detect when someone is trying to uninstall or disable the product, and then cut off Internet access altogether until it has been properly reconfigured. Filtering companies actively take steps to make sure that children are not able to come up with ways to circumvent their filters. Filtering companies monitor the Web to identify any methods for circumventing filters, and when such methods are found, the filtering companies respond by putting in extra protections in an attempt to make sure that those methods do not succeed with their products. Id. at 795 (citations omitted). The court also found that “[i]t is difficult for children to circumvent filters because of the technical ability and expertise necessary to do so.... ” Id. Finally, the court found that “filters generally block about 95% of sexually explicit material.” Id. After describing filtering technology, the District Court concluded that the Government “failed to successfully defend against the plaintiffs’ assertion that filter software and the Government’s promotion and support thereof is a less restrictive alternative to COPA.” Id. at 813. The court reasoned that “unlike COPA there are no fines or prison sentences associated with filters which would chill speech. Also unlike COPA, ... filters are fully customizable and may be set for different ages and for different categories of speech or may be disabled altogether for adult use. As a result, filters are less restrictive than COPA.” Id. (citations omitted). The District Court also concluded that the Government “failed to show that filters are not at least as effective as COPA at *202 protecting minors from harmful material on the Web.” Id. at 814. The court determined that COPA will not reach sexually explicit materials on the Web that originate from foreign sources, its affirmative defenses are not effective, and it is unlikely that COPA will be enforced widely. The court found that: filters block sexually explicit foreign material on the Web, parents can customize filter settings depending on the ages of their children and what type of content they find objectionable, and filters are fairly easy to install and use.... [Filters are very effective at blocking potentially harmful sexually explicit materials. Id. at 815 (citations omitted). The court concluded that “[e]ven defendant’s own study shows that all but the worst performing filters are far more effective than COPA would be at protecting children from' sexually explicit material on the Web....” Id. The Government does not challenge the District Court’s factual findings and therefore we need not set forth the evidence on which the court based its findings. The Government does contend, however, that the District Court erred in concluding that filters are a less restrictive alternative because the court applied a “flawed analytical framework” and that filters cannot be considered a less restrictive alternative because they are part of the “status quo.” Appellant’s Br. at 43-44. But the Supreme Court’s statement on this issue contravenes the Government’s argument: In considering this question, a court assumes that certain protected speech may be regulated, and then asks what is the least restrictive alternative that can be used to achieve that goal.... The purpose of the test is to ensure that the speech is restricted no further than necessary to achieve the goal, for it is important to assure that legitimate speech is not chilled or punished. For that reason, the test does not begin with the status quo of existing regulations, then ask whether the challenged restriction has some additional ability to achieve Congress’ legitimate interest. Any restriction on speech could be justified under that analysis. Instead, the court should ask whether the challenged regulation is the least restrictive means among available, effective alternatives. Ashcroft, 542 U.S. at 666, 124 S.Ct. at 2791. This reasoning explains why the Court then instructed the parties to update the factual record regarding “the effectiveness of filtering software” so that the District Court could determine whether “filters are less effective than COPA.” Id. at 671, 124 S.Ct. at 2794. Accordingly, the Government is incorrect in its assertion that the District Court applied an erroneous analytical framework. We agree with the District Court’s conclusion that filters and the Government’s promotion of filters are more effective than COPA. The Supreme Court already has written how the Government could act to promote and support the use of filters: Congress undoubtedly may act to encourage the use of filters. We have held that Congress can give strong incentives to schools and libraries to use them. It could also take steps to promote their development by industry, and their use by parents. It is incorrect, for that reason, to say that filters are part of the current regulatory status quo. The need for parental cooperation does not automatically disqualify a proposed less restrictive alternative. In enacting COPA, Congress said its goal was to prevent the ‘widespread availability of the Internet’ from providing ‘opportunities for minors to access materials through the World Wide Web in a man *203 ner that can frustrate parental supervision or control.’ COPA presumes that parents lack the ability, not the will, to monitor what their children see. By enacting programs to promote use of filtering software, Congress could give parents that ability without subjecting protected speech to severe penalties. Id. at 669-70, 124 S.Ct. at 2793 (citations omitted). As the District Court pointed out, filters can be used to block foreign Web sites, which COPA does not regulate. Though the Government contends that COPA applies to foreign Web sites, the Supreme Court already has rejected the Government’s construction of the statute. In Ashcroft the Court stated that: a filter can prevent minors from seeing all pornography, not just pornography posted to the Web from America.... COPA does not prevent minors from having access to those foreign harmful materials.... [I]f COPA is upheld, ... providers of the materials that would be covered by the statute simply can move their operations overseas. Id. at 667, 124 S.Ct. at 2792. In light of the Supreme Court’s express conclusion that COPA does not apply to foreign Web sites — a determination that does not depend upon the facts developed at the later trial in the District Court — we cannot construe COPA to apply to foreign Web sites. Given the vast quantity of speech that COPA does not cover but that filters do cover, it is apparent that filters are more effective in advancing Congress’s interest, as it made plain it is in COPA. Moreover, filters are more flexible than COPA because parents can tailor them to their own values and needs and to the age and maturity of their children and thus use an appropriate flexible approach differing from COPA’s “one size fits all” approach. Finally, the evidence makes clear that, although not flawless, with proper use filters are highly effective in preventing minors from accessing sexually explicit material on the Web. At oral argument, the Government made much of a study that found that only 54 percent of parents use filters. But the Government has neglected the fact that this figure represents a 65 percent increase from a prior study done four years earlier, which indicates that significantly more families are using filters. App. at 159-60. Furthermore, the circumstance that some parents choose not to use filters does not mean that filters are not an effective alternative to COPA. Though we recognize that some of those parents may be indifferent to what their children see, others may have decided to use other methods to protect their children — such as by placing the family computer in the living room, instead of their children’s bedroom — or trust that their children will voluntarily avoid harmful material on the Internet. Studies have shown that the primary reason that parents do not use filters is that they think they are unnecessary because they trust their children and do not see a need to block content. Id. at 160, 164, 278, 1567. The Government simply has not carried its burden of showing that COPA is a more effective method than filters in advancing the Government’s compelling interest as evidenced in COPA. In addition to being more effective, it is clear that filters are less restrictive than COPA. As the Supreme Court has stated: [fjilters are less restrictive than COPA. They impose selective restrictions on speech at the receiving end, not universal restrictions at the source. Under a filtering regime, adults without children may gain access to speech they have a right to see without having to identify themselves or provide their credit card *204 information. Even adults with children may obtain access to the same speech on the same terms simply by turning off the filter on their home computers. Above all, promoting the use of filters does not condemn as criminal any category of speech, and so the potential chilling effect is eliminated, or at least much diminished. All of these things are true, moreover, regardless of how broadly or narrowly the definitions in COPA are construed. Ashcroft, 542 U.S. at 667, 124 S.Ct. at 2792. Although the Supreme Court made this statement after reviewing the record from the hearing on the preliminary injunction, the evidence produced at the trial on the merits confirms the Court’s initial impression. Unlike COPA, filters permit adults to determine if and when they want to use them and do not subject speakers to criminal or civil penalties. During oral argument, the Government contended that the First Amendment does not prohibit Congress from adopting a “belt-and-suspenders” approach to addressing the compelling government interest of protecting minors from accessing harmful material on the Web, with filters acting as the “belt” and COPA as the “suspenders.” But as counsel for plaintiffs correctly pointed out, under the First Amendment, if the belt works at least as effectively as the suspenders, then the Government cannot prosecute people for not wearing suspenders. Here, based on the prior litigation in the Supreme Court and this Court in ACLU II and the District Court’s findings on the remand, the Government has not shown that COPA is a more effective and less restrictive alternative to the use of filters and the Government’s promotion of them in effectuating COPA’s purposes. Indeed, we would reach this conclusion on the basis of either the prior litigation or the District Court’s findings on the remand. Accordingly, COPA fails the third prong of a strict scrutiny analysis and is unconstitutional. C. Vagueness and Overbreadth The Government also challenges the District Court’s decision that COPA facially violates the First and Fifth Amendments because it is impermissibly vague and overbroad. 1. Vagueness The Supreme Court recently described the vagueness doctrine: Vagueness doctrine is an outgrowth not of the First Amendment, but of the Due Process Clause of the Fifth Amendment. A conviction fails to comport with due process if the statute under which it is obtained fails to provide a person of ordinary intelligence fair notice of what is prohibited, or is so standardless that it authorizes or encourages seriously discriminatory enforcement. Although ordinarily a plaintiff who engages in some conduct that is clearly proscribed cannot complain of the vagueness of the law as applied to the conduct of others, we have relaxed that requirement in the First Amendment context, permitting plaintiffs to argue that a statute is overbroad because it is unclear whether it regulates a substantial amount of protected speech. But perfect clarity and precise guidance have never been required even of regulations that restrict expressive activity. Williams, 128 S.Ct. at 1845 (citations, quotation marks, and brackets omitted). The Court further explained: What renders a statute vague is not the possibility that it will sometimes be difficult to determine whether the incriminating fact it establishes has been proved; but rather the indeterminacy of precisely what that fact is. Thus, we have *205 struck down statutes that tied criminal culpability to whether the defendant’s conduct was ‘annoying’ or ‘indecent’— wholly subjective judgments without statutory definitions, narrowing context, or settled legal meanings. Id. at 1846. Our discussion in ACLU II of the question of whether COPA is impermissi-bly vague was quite limited but in a footnote we stated that we considered COPA’s use of the term “minor” as incorporated in COPA’s definition of “material that is harmful to minors” to be impermissibly vague. We reached this conclusion because we believed that “a Web publisher will be forced to guess at the bottom end of the range of ages to which the statute applies,” and thus will not have “fair notice of what conduct would subject them to criminal sanctions under COPA” and “will be deterred from engaging in a wide range of constitutionally protected speech.” 322 F.3d at 268 n. 37. The District Court on the remand concluded that COPA is vague for several reasons. First, the court pointed out that COPA utilizes two different scienter requirements — “knowingly” and “intentionally” — but does not define either standard. Gonzales, 478 F.Supp.2d at 816-17. Second, the court determined that although Congress intended COPA to apply solely to commercial pornographers, the phrase “communication for commercial purposes” as modified by the phrase “engaged in the business” does not limit COPA’s application to commercial pornographers. Id. at 817. Thus, Web publishers that are not commercial pornographers will be uncertain as to whether they will face prosecution under the statute, chilling their speech. Id. Third, the court found that the definition of “minor” as any person under 17 years of age creates vagueness in COPA because materials that could have “serious literary, artistic, political, or scientific value” for a 16-year-old would not necessarily have the same value for a three-year-old. Id. Thus, Web publishers cannot tell which of these minors should be considered in deciding the content of their Web sites. Id. at 817-18. Fourth, the court stated that COPA’s use of the phrase “as a whole” is vague because it is unclear how that phrase would apply to the Web. Id. at 818. The Government contends that the District Court erred in finding COPA imper-missibly vague and argues that the statutory provisions that the District Court concluded rendered the statute vague instead served to limit the reach of the statute. We are bound by our conclusion in ACLU II that COPA’s definition of “minor” renders the statute vague. Furthermore we agree with the District Court’s conclusion that COPA’s use of the phrases and terms “communication for commercial purposes,” “as a whole,” “intentional,” and “knowing” renders it vague, for the reasons the District Court stated in its opinion. 2. Overbreadth The Supreme Court also addressed the First Amendment overbreadth doctrine in Williams, stating that: [A] statute is facially invalid if it prohibits a substantial amount of protected speech. The doctrine seeks to strike a balance between competing social costs. On the one hand, the threat of enforcement of an overbroad law deters people from engaging in constitutionally protected speech, inhibiting the free exchange of ideas. On the other hand, invalidating a law that in some of its applications is perfectly constitutional— particularly a law directed at conduct so antisocial that it has been made crimi *206 nal — has obvious harmful effects. In order to maintain an appropriate balance, we have vigorously enforced the requirement that a statute’s overbreadth be substantial, not only in an absolute sense, but also relative to the statute’s plainly legitimate sweep. Invalidation for overbreadth is strong medicine that is not to be casually employed. 128 S.Ct. at 1838 (citations and quotation marks omitted). In ACLU II we held that COPA is “substantially overbroad” because: it places significant burdens on Web publishers’ communication of speech that is constitutionally protected as to adults and adults’ ability to access such speech. In so doing, COPA encroaches upon a significant amount of protected speech beyond that which the Government may target constitutionally in preventing children’s exposure to material that is obscene for minors. 322 F.3d at 266-67. We found that COPA’s definition of “material harmful to minors” “impermissibly places at risk a wide spectrum of speech that is constitutionally protected” because it “calls for evaluation of ‘any material’ on the Web in isolation.” Id. at 267. Thus, we explained: an isolated item located somewhere on a Web site that meets the ‘harmful to minors’ definition can subject the publisher of the site to liability under COPA, even though the entire Web page (or Web site) that provides the context for the item would be constitutionally protected for adults (and indeed, may be protected as to minors). Id. We also found that COPA’s definition of “minors” renders the statute overinclu-sive because it “broadens the reach of ‘material that is harmful to minors’ under the statute to encompass a vast array of speech that is clearly protected for adults — and indeed, may not be obscene as to older minors....” Id. at 268. We next found that COPA’s definition of “commercial purposes” rendered the statute over-broad for the same reasons that it failed strict scrutiny. Id. at 269. We also found that “COPA’s application of ‘community standards’ exacerbates these constitutional problems in that it further widens the spectrum of protected speech that COPA affects.” Id. at 270. We stated that “COPA essentially requires that every Web publisher subject to the statute abide by the most restrictive and conservative state’s community standards in order to avoid criminal liability.” Id. (quoting ACLU I, 217 F.3d at 166). Finally, we found that there was no available narrowing construction that would make COPA constitutional. Id. at 270-71. These conclusions bind us here. The District Court also found that COPA is overbroad for several reasons. First, the court determined that the vagueness of the phrases “communication for commercial purposes” and “engaged in the business” means that COPA could apply to a wide swath of the Web and thus COPA would prohibit and chill a substantial amount of constitutionally protected speech for adults. Gonzales, 478 F.Supp.2d at 819. Second, because the definition of “minor” includes any person under 17, Web publishers do not have fair notice regarding what they can place on the Web that will not be considered harmful to any minor. Id. Thus, the definition of “minor” renders COPA overinclusive because it broadens the statute to encompass a large array of protected speech. Id. Finally, the court found that because the statute does not reference commercial pornographers, it found that it could not read such a limitation into the statute to save it from being overbroad. Id. at 819-20. *207 The Government claims that COPA is not overbroad, but it is clear that our prior decision in ACLU II binds us on this issue. It is apparent that COPA, like the Communications Decency Act before it, “effectively suppresses a large amount of speech that adults have a constitutional right to receive and to address to one another,” Reno, 521 U.S. at 874, 117 S.Ct. at 2346, and thus is overbroad. For this reason, COPA violates the First Amendment. V. CONCLUSION In sum, COPA cannot withstand a strict scrutiny, vagueness, or overbreadth analysis and thus is unconstitutional. We reach our result both through the application of the law-of-the-case doctrine to our determination in ACLU II and on the basis of our independent analysis of COPA and would reach the same result on either basis standing alone. For the foregoing reasons, we will affirm the District Court’s March 22, 2007 order. Footnotes [1] The Government, however, does not challenge the District Court’s determination that plaintiffs have standing to bring this action. [2] The Government wrote this letter in response to our request that the parties file supplemental letter briefs on the law-of-the-case issue. [3] Our opinion in ACLU II is not entirely clear on this point. We started our discussion of the least restrictive alternative question by indicating that "[w]e are also satisfied that COPA does not employ the 'least restrictive means' to effect the Government’s compelling interest in protecting minors.” ACLU II, 322 F.3d at 261. Then in considering that question in more detail we discussed filters at length. At one point in the opinion we stated that "filtering software is a less restrictive alternative that can allow parents some measure of control over their children’s access to speech that parents consider inappropriate.” Id. at 263. At several other points, we also stated that COPA is not the least restrictive alternative. See id. at 261 ("We are ... satisfied that COPA does not employ the least restrictive means’ to effect the Government's compelling interest in protecting minors.”); id. at 265-66 ("The existence of less restrictive alternatives renders COPA unconstitutional under strict scrutiny.... COPA also fails strict scrutiny because it does not use the least restrictive means to achieve its ends.... Congress could have, but failed to employ the least restrictive means to accomplish its legitimate goal...Nevertheless we stated that "[w]e agree with the District Court that the various blocking and filtering techniques which that Court discussed may be substantially less restrictive than COPA in achieving COPA's objective of preventing a minor’s access to harmful material.” Id. at 265 (emphasis added). Because of this statement, we cannot state with certainty that ACLU II squarely holds that filters are less restrictive than COPA, though it probably does. Thus, for law-of-the-case purposes, we might not consider ourselves bound on this appeal by that determination. Of course, this discussion of whether we determined that filters are less restrictive than COPA or that filters only may be less restrictive than COPA is somewhat academic, for on the appeal of ACLU II the Supreme Court explicitly addressed this issue and, though remanding the case, flatly indicated that filters are "less restrictive” than COPA, Ashcroft, 542 U.S. at 667, 124 S.Ct. at 2792, and that Court's conclusions supersede our decision in ACLU II on this point.