The plaintiff, an editor, publisher, and part-owner of a newspaper distributed exclusively through electronic means, brings this First Amendment challenge to § 223(d) of the Communications Decency Act of 1996 ("CDA") criminalizing the use of interactive computer services to display "patently offensive" sexually explicit material such that it is available to persons under the age of eighteen. The plaintiff seeks a preliminary injunction barring application of the section. The three-judge panel, appointed pursuant to 28 U.S.C. § 2284, held that: (1) plaintiff has not sustained his burden of demonstrating a likelihood of success on his claim that § 223(d) is unconstitutionally vague, but that (2) the plaintiff has demonstrated a likelihood of success on his claim that § 223(d) is unconstitutionally overbroad in that it bans protected indecent communication between adults. On this second point, the court concluded that most content providers' ability to comply with the requirements of the affirmative defenses set out in the statute depends on the actions of third parties, such as software manufacturers, whose cooperation is not required under the statute or otherwise mandated. The technological impossibility of independent compliance with the affirmative defenses renders § 223(d) unconstitutional as an overbroad prohibition on constitutionally protected indecent speech between adults.
JOSE A. CABRANES, Circuit Judge :
We address here the constitutionality of a provision of the Communications Decency Act of 1996 ("CDA") with an undeniably worthy goal: to limit the exposure of children to sexually explicit, though not legally obscene, materials available "on line"--that is, capable of being displayed and "accessed" by increasingly common interactive computer services. 47 U.S.C. § 223(d), as added by the CDA on February 8, 1996, criminalizes the use of an interactive computer service to display, in a manner available to persons under eighteen, sexually explicit material that is "patently offensive" by contemporary community standards. Plaintiff Joe Shea, the editor, publisher, and part-owner of a newspaper distributed solely by electronic means, filed this action on February 8, 1996, claiming that § 223(d) is (1) void for vagueness, in that it fails to give ordinary citizens sufficient notice of what conduct will subject them to prosecution or criminal liability; and (2) substantially overbroad, in that it targets a broader category of speech than necessary to achieve the government's goal and constitutes a ban on certain constitutionally protected speech between adults.
As editor of an on-line newspaper, the plaintiff is one of a growing number of citizens who employ an array of widely accessible and constantly evolving media technologies to gather and disseminate information and ideas. In passing the CDA, Congress explicitly recognized that these technologies foster "true diversity of political discourse, unique opportunities for cultural development, and myriad avenues for intellectual activity." Pub. L. No. 104-104, § 509(a)(3), 110 Stat. 56, 138 (1996) (to be codified at 47 U.S.C. § 230(a)(3)). The range of tools and forums available for users of interactive computer services is astounding: with access to the web of computer networks known as the Internet, a scholar can contact a distant computer and make use of its capabilities; a researcher can peruse the card catalogs of libraries across the globe; users around the world can debate politics, sports, music, and literature. However trivial some of their uses might seem, emerging media technologies quite simply offer an unprecedented number of individual citizens an opportunity to speak and to be heard--at very little cost--by audiences around the world. In that sense, we are encountering a communications medium unlike any we have ever known.
In an attempt to limit the availability of certain materials in interactive computer services, Congress enacted a statute of unprecedented sweep: the new § 223(d) purports to regulate not only how commercial purveyors of obscene or pornographic materials may advertise and sell their products on line, but also how private individuals who choose to exchange certain constitutionally protected communications with one another can do so. The question presented is whether our Constitution tolerates this level of governmental intrusion into how adults speak to one another.
We conclude, first, that the plaintiff has not sustained his burden of demonstrating a likelihood of success on his claim that § 223(d) is unconstitutionally vague. The definition of material regulated by this section is a familiar one, repeatedly upheld against vagueness challenges in a line of jurisprudence concerning television and radio broadcasting, cable programming, and commercial telephone services. We do, however, conclude that the plaintiff has demonstrated a likelihood of success on his overbreadth claim, that § 223(d) would serve as a ban on constitutionally protected indecent communication between adults. The Government concedes that strict scrutiny is appropriately applied to this claim and that § 223(d) would, on its own, act as an unconstitutional total ban on indecent communication, protected and unprotected alike, but argues that two affirmative defenses set out in § 223(e)(5) serve to shield adults engaging in constitutionally protected indecent communication from criminal liability.
The evidentiary record in this case compels the conclusion that, given the current state of technology, most adult content providers wishing to engage in constitutionally protected indecent speech will be unable to avail themselves of these affirmative defenses. Only a limited subset of on-line content providers, commercial providers on the World Wide Web, can avail themselves of the defense set out in § 223(e)(5)(B), leaving both non-commercial providers of Web content and content providers using all other modes of on-line communication unprotected. The evidence further demonstrates that content providers' ability to comply with the terms of the second defense--the so-called good-faith defense--depends on the actions of third parties, such as software manufacturers, whose cooperation is not required under the CDA or otherwise mandated. There is no feasible means, with our current technology, for someone to provide indecent content on line with any certainty that even his best efforts at shielding the material from minors will be "effective," as the language of the good-faith defense requires.
Because neither of the affirmative defenses set out in § 223(e)(5) can, with our current technology, effectively protect adult content providers wishing to engage in constitutionally protected indecent communication, we reach the inescapable conclusion that § 223(d) will serve to chill protected speech. We therefore find that the plaintiff has demonstrated a likelihood of success on the merits of his claim that § 223(d) is unconstitutionally overbroad.
We are mindful of our obligation to construe a federal statute to avoid constitutional problems if it is possible to do so, but we are equally mindful of the limits of the judicial power under our Constitution and we decline the Government's invitation to perform radical surgery on a statute dealing with a difficult problem in a rapidly changing area of technology; in sum, we respectfully decline the invitation to legislate from the bench.
In setting aside the challenged provisions, we do not question the legitimacy of the government's interest in safeguarding children from exposure to certain materials available on line nor suggest that other legislation on another day, carefully tailored to technological realities, may not pass constitutional muster. We also do not consider, nor attempt to delineate, the range of circumstances, if any, in which Congress could now or in the future constitutionally impose content-based restrictions upon communications in the developing medium we explore here.
Plaintiff Joe Shea is the editor-in-chief, part-owner, and publisher of the American Reporter, a daily newspaper distributed solely by electronic means. On February 8, 1996, following the signing of the Telecommunications Act of 1996, Pub. L. No. 104-104, 110 Stat. 56, the American Reporter published an editorial, (Complaint, Ex. 1) criticizing Title V of the Act, known as the Communications Decency Act of 1996 ("CDA"). The editorial contained language arguably falling within the scope of a provision of the CDA criminalizing the transmission or display of certain content in a manner available to minors:
(A) uses an interactive computer service to send to a specific person or persons under 18 years of age, or
(B) uses any interactive computer service to display in a manner available to a person under 18 years of age,
any comment, request, suggestion, proposal, image, or other communication that, in context, depicts or describes, in terms patently offensive as measured by contemporary community standards, sexual or excretory activities or organs, regardless of whether the user of such service placed the call or initiated the communication; . . .
shall be fined under title 18, United States Code, or imprisoned not more than two years, or both.
Pub. L. No. 104-104, § 502(2)(d), 106 Stat. at 133 (to be codified at 47 U.S.C. § 223(d)). Also on February 8, the plaintiff filed this action on behalf of the American Reporter, seeking a declaration that 47 U.S.C. § 223(d) is unconstitutionally overbroad and vague. On February 17, the plaintiff moved for preliminary injunctive relief to prevent the Department of Justice from enforcing the provision in question.
Pursuant to § 561(c) of the Telecommunications Act and 28 U.S.C. § 2284, the Chief Judge of the United States Court of Appeals for the Second Circuit designated this three-judge district court to consider the plaintiff's facial challenge to the constitutionality of § 223(d). The Court heard opening arguments on April 3, 1996. Because the plaintiff's facial challenge to § 223(d) raised the question of whether Internet users can, under current technology, meet the requirements for certain defenses provided for in the Act, the Court concluded that an evidentiary hearing would be appropriate. In light of the pendency of consolidated proceedings for preliminary injunctive relief before a three-judge court in the Eastern District of Pennsylvania raising, among other claims, a facial challenge to § 223(d), see Complaint, American Civil Liberties Union v. Reno, 1996 U.S. Dist. LEXIS 1617, No. 96-963 (E.D. Pa. filed Feb. 8, 1996); Complaint, American Library Ass'n v. U.S. Dep't of Justice, No. 96-1458 (E.D. Pa. filed Feb. 27, 1996) (collectively "the Philadelphia litigation"), the Court directed the parties to consider methods of easing the fact-finding process by entering into a range of stipulations regarding the current state of technology and incorporating relevant portions of the record in the Philadelphia litigation. The Court received additional demonstrative and testimonial evidence on April 29, April 30, and May 6, and heard closing arguments on June 3, 1996. Following supplemental briefing by the parties, the plaintiff's motion for preliminary injunctive relief was submitted for decision on June 13, 1996.
On June 11, 1996, the three-judge court in the Philadelphia litigation concluded, inter alia, that the provision of the CDA here challenged by the plaintiff does not withstand constitutional scrutiny. American Civil Liberties Union v. Reno, 929 F. Supp. 824, 1996 WL 311865, at *27 (E.D. Pa. 1996) (" ACLU/ALA "). All three judges agreed that the CDA is substantially overbroad, in that it effectively forces many Internet users (specifically, non-commercial, not-for-profit entities and "even many commercial organizations") to forgo constitutionally protected speech or risk criminal prosecution. Id. at *32-*33 (Sloviter, C.J.); id. at *37 (Buckwalter, J.); id. at *49 (Dalzell, J.). Additionally, two of the judges concluded that § 223(d)'s definition of covered speech is unconstitutionally vague. Id. at *36 (Sloviter, C.J.); id. at *37 (Buckwalter, J.). The decision in the Philadelphia litigation does not preclude this Court from deciding the issues presented,
to which we now turn.
We enter the following findings of fact, many of which are undisputed, the subject of stipulations by the parties, or submitted by the defendant and adopted by us, pursuant to Rule 52(a) of the Federal Rules of Civil Procedure. Although we here consider a so-called facial challenge to a statute, we deemed it appropriate and necessary in the unusual circumstances presented here, and a reasonable exercise of our discretion, to establish a basic record of the facts regarding the new and evolving communications media that is the subject of this legislation.
Section 223(d) targets the use of an "interactive computer service" to send or display patently offensive materials. Although § 223 itself contains no definition of that term, the definition applicable to the new 47 U.S.C. § 230--also added by the CDA--makes clear that the term encompasses means of making "content"
available to multiple users both on the vast web of linked networks popularly known as "the Internet" and on other information systems (such as electronic bulletin boards maintained by educational institutions or nonprofit organizations) not physically linked to the Internet. See Pub. L. No. 104-104, § 509(e)(2), 110 Stat. at 139 (to be codified at 47 U.S.C. § 230(e)(2)). We draw upon the stipulations of the parties and the testimony adduced at the three-day evidentiary hearing to describe: (1) the nature of the medium targeted by § 223(d), focusing in part on the degree of control that those who transmit content have over who will receive it; (2) the availability of certain categories of potentially objectionable material on line; (3) the development of software and labeling standards enabling parents to limit their children's exposure to objectionable on-line content; and (4) the potential for tagging and verification procedures that content providers can use in an effort to shield minors from sexually explicit content that they provide.
As we do so, we unavoidably--and with apologies to all others with a similar aversion to "cyberspeak"--adopt some of the terminology that has developed in conjunction with this technology. We endeavor, to the extent possible, to avoid the jargon of this field, and to define our terms wherever possible, for the sake of the clarity of this record and this opinion, as well as for the benefit of any reader required to review our work.
A. The Development of the Internet
Although "the Internet" now formally describes a collection of more than 50,000 networks linking some nine million host computers in ninety countries, it has existed for nearly three decades on a much smaller scale. What we now refer to as the Internet grew out of an experimental project of the Department of Defense's Advanced Research Projects Administration ("ARPA") designed to provide researchers with direct access to supercomputers at a few key laboratories and to facilitate the reliable transmission of vital communications. (Declaration of William J. Hoffman ("Hoffman Decl."), Ex. 4, at 11-12) ARPA supplied funds to link computers operated by the military, defense contractors, and universities conducting defense-related research through dedicated phone lines, creating a "network" known as ARPANet. (Parties' Stipulations in Preparation for Preliminary Injunction Hearing ("Joint Stip.") PP 6-7; Hoffman Decl., Ex. 3, at 3; id. Ex. 4, at 11) Programs on the linked computers implemented a technical scheme known as "packet-switching," through which a message from one computer to another would be subdivided into smaller, separately addressed pieces of data, known as "packets," sent independently to the message's destination and reassembled upon arrival. (Joint Stip. P 9) Each computer on the network was in turn linked to several other computers, creating any number of routes that a communication from one computer could follow to reach its destination. If part of the network were damaged, a portion of the message could be re-routed automatically over any other path to its ultimate destination, a characteristic of the network intended initially to preserve its operability in the event of enemy attack. (Id. PP 7-8; Hoffman Decl., Ex. 3, at 3; id. Ex. 4, at 12)
Having successfully implemented a system for the reliable transfer of information over a computer network, ARPA began to support the development of communications protocols for transferring data between different types of computer networks. Universities, research facilities, and commercial entities began to develop and link together their own networks implementing these protocols; these networks included a high-speed "backbone" network known as NSFNet, sponsored by the National Science Foundation, smaller regional networks, and, eventually, large commercial networks run by organizations such as Sprint, IBM, and Performance Systems International (commonly known as "PSI"). (Hoffman Decl., Ex. 3, at 3; id. Ex. 4, at 13-14) As faster networks developed, most network traffic shifted away from ARPANet, which formally ceased operations in 1990. (Id. Ex. 3, at 3) What we know as "the Internet" today is the series of linked, overlapping networks that gradually supplanted ARPANet. Because the Internet links together independent networks that merely use the same data transfer protocols, it cannot be said that any single entity or group of entities controls, or can control, the content made publicly available on the Internet or limits, or can limit, the ability of others to access public content. Rather, the resources available to one with Internet access are located on individual computers around the world. (Joint Stip. P 11)
It is estimated that as many as forty million individuals have access to the information and tools of the Internet, and that figure is expected to grow to 200 million by the year 1999. (Id. P 3) Access to the Internet can take any one of several forms. First, many educational institutions, businesses, libraries, and individual communities maintain a computer network linked directly to the Internet and issue account numbers and passwords enabling users to gain access to the network directly or by modem.
(Id. PP 12-14) Second, "Internet service providers," generally commercial entities charging a monthly fee, offer modem access to computers or networks linked directly to the Internet. (Id. P 16) Third, national commercial "on-line services"--such as America Online, CompuServe, Prodigy, and Microsoft Network--allow subscribers to gain access to the Internet while providing extensive content within their own proprietary networks. (Id. P 17) Finally, organizations and businesses can offer access to electronic bulletin-board systems--which, like national on-line services, provide certain proprietary content; some bulletin-board systems in turn offer users links to the Internet. (Id. P 18)
B. Categories of Internet Use
For our purposes, there are two loose and overlapping categories of Internet use. First, an individual who has secured access to the Internet can correspond or exchange views with one or many other Internet users. Second, a user can locate and retrieve information available on other computers. We explore these categories in greater detail below. As will become clear, distinctions in how Internet content is transmitted affect the degree of control that providers of content have over who will be able to gain access to their communications;
we will return to the legal significance of these distinctions at a later juncture. For any communication to take place over the Internet, two pieces of software,
adhering to the same communications protocol, are required. A user must have access to certain kinds of "client" software, which enables his computer to communicate with and make requests of remote computers where information is stored; these remote computers must be running "server" software, which provides information in response to requests by client software. (Declaration of Dr. Dan R. Olsen, Jr. ("Olsen Decl."), PP 13-14)
1. Communicating with Other Internet Users
Perhaps the most widely used Internet service is electronic mail, or "e-mail." Using any one of dozens of available "mailers"--client software capable of reading and writing e-mail--a user is able to address and transmit a message to one or more specific individuals. (Joint Stip. P 21) A user can also "subscribe" to an electronic mailing list on a topic of interest; the user receives a copy of messages posted by other subscribers and, in turn, can post messages for forwarding to the full mailing list. Once a mailing list is established, it is typically maintained using a "mail exploder"--a program such as "listserv" running on the server on which the list resides--that automatically (i.e., without human intervention) responds to a user's request to be added to or removed from the list of subscribers and retransmits messages posted by a subscriber to others on the mailing list. (Id. P 22) Some mailing lists are "closed": a user's request to join the list requires the approval of an individual who maintains the list. (Id.) Mailing lists (both open and closed) may also be "moderated": all messages posted to the list are forwarded to a moderator, who approves certain messages and retransmits them to subscribers. (Id.) An individual sending a message that will be retransmitted by a mail exploder program has no way of knowing the e-mail addresses of other subscribers. (Olsen Decl. P 19; Testimony of Gordon C. Galligher, Jr., Tr. at 181) Even if the user could obtain an e-mail address for each subscriber to a particular list, those addresses alone would provide no authoritative information about subscribers. There is no directory that identifies persons using a certain e-mail address. In addition, a user can avoid disclosing his true e-mail address by developing an e-mail "alias" or by using an "anonymous remailer"--a server that purges identifying information from a communication before forwarding it to its destination. (Defendant's Response to Plaintiff's Request for Admissions ("Defendant's Adm.") No. 22; Galligher Test. at 173)
Internet users may also transmit or receive "articles" posted daily to thousands of discussion groups, arranged by subject matter and known as "newsgroups," available through an electronic bulletin-board system known as "Usenet." When a user with access to a Usenet server--that is, a computer participating in the Usenet system--posts an article to a particular newsgroup, the server automatically forwards the article to adjacent Usenet servers, which in turn forward it to other servers, until the article is available on all Usenet sites that furnish access to the newsgroup in question. (Joint Stip. P 23) Once a message reaches a particular Usenet site, it is temporarily stored there so that individual users--running client software, known as a "newsreader," capable of sorting articles according to header information identifying the newsgroup to which the article was posted--can review and respond to the message. (Id. ; Hoffman Decl., Ex. 4, at 129) Some Usenet newsgroups are moderated; messages to the newsgroup are forwarded to an individual who selects those appropriate for distribution. (Joint Stip. P 23) Because Usenet articles are distributed to (and made available on) multiple servers, one who posts an article to a newsgroup has no way of knowing who will choose to retrieve it, whether or not the newsgroup is moderated. (Galligher Test., Tr. at 170, 174-75) There is no newsgroup equivalent of a "closed" mailing list: access to a particular newsgroup can only be limited by restricting the number of servers participating in the newsgroup. (Testimony of Clay Shirky, Tr. at 251)
The Internet also offers opportunities for multiple users to interact in real time. Using a program called "Talk," two users can exchange messages while they are both on line; a message typed on one user's computer will appear almost immediately on the other's screen. (Joint Stip. P 25) Servers running so-called "chat" software, such as Internet Relay Chat ("IRC"), permit multiple users to converse by selecting one of many discussion "channels" active at any time. Commercial on-line services such as America Online, CompuServe, Prodigy, and the Microsoft Network offer their own chat systems for their members. (Id. P 26) Having joined a channel, the user can see and read messages transmitted by other users, each identified by a name the user selects upon joining the channel. (Id. P 25) Individual participants in IRC discussions know other participants only by the names they choose upon entering the discussion; users can participate anonymously by using a pseudonym.
2. Locating and Retrieving Information on the Internet
Individuals with Internet access can take advantage of a number of tools for locating and retrieving information and resources stored on remote computers. One who wishes to make certain articles, files, or software available to other users will set up a server, adhering to certain communications protocols, capable of retrieving and presenting stored information in response to a request from client software using the same communications protocol. (Olsen Decl. PP 13, 16; Galligher Test., Tr. at 131)
a. File-Transfer Protocol ("FTP")
One type of software implements a set of conventions for copying files from a host computer known as "file-transfer protocol" ("FTP"). With appropriate client software, a user with an account on the host computer can contact the server, view a directory of available files, and copy one or more of those files to his own computer. In addition to making files available to users with accounts, thousands of content providers also make files available for "anonymous" retrieval by users who do not possess an account on the host computer.
(Hoffman Decl., Ex. 3, at 1-2, 5; id. Ex. 4, at 187; Joint Stip. P 29) A content provider who makes files available for retrieval by anonymous FTP has no way of discerning who gains access to the files.
The third and perhaps best known method of locating and accessing information on the Internet is by exploring the World Wide Web. Documents available on the Web are not collected in any central location; rather, they are stored on servers around the world running Web server software. (Joint Stip. PP 31, 38, 40) To gain access to the content available on the Web, a user must have a Web "browser"--client software, such as Netscape Navigator, Mosaic, or Internet Explorer, capable of displaying documents formatted in "hypertext markup language" ("HTML"), the standard Web formatting language. (Galligher Test., Tr. at 125; Joint Stip. PP 31, 43) Each document has an address, known as a Uniform Resource Locator ("URL"), identifying, among other things, the server on which it resides; most documents also contain "links"--highlighted text or images that, when selected by the user, permit him to view another, related Web document. (Joint Stip. P 34) Because Web servers are linked to the Internet through a common communications protocol, known as hypertext transfer protocol ("HTTP"), a user can move seamlessly between documents, regardless of their location; when a user viewing a ...