The First Amendment Does the Work: The One Test Answer for American Internet Law

American internet law is built from court opinions more than from statutes. The First Amendment does more work in the field than any enacted text, because the single liability provision Congress wrote for the network has been applied in thousands of lawsuits without ever receiving an authoritative construction from the Supreme Court. The rules that determine what governments may require of platforms, what states may forbid on the network, and what pressure officials may apply to intermediaries come from five lines of Supreme Court decisions that run from Reno v. American Civil Liberties Union in 1997 to the paired content moderation rulings of 2024.

The first line begins with Reno itself. In striking down the indecency provisions of the Communications Decency Act, the Court held that expression on the internet receives the full First Amendment protection given to print rather than the diminished protection tolerated for broadcast. That choice of analogy set the constitutional default for everything that followed.

The second line distinguishes between the instruments Congress may use. The funding condition line holds that Congress may attach conditions to money it need not spend, while direct prohibitions on protected expression face strict scrutiny. United States v. American Library Association sustained the Children’s Internet Protection Act as a condition on a federal subsidy, while Ashcroft v. American Civil Liberties Union kept the criminal Child Online Protection Act enjoined until it expired without ever being enforced.

The third line is the access line. Packingham v. North Carolina held that a state may not make social networking platforms off-limits to an entire class of people, treating the ability to be present where the exchange of views happens as constitutionally significant in its own right. The decision separated regulating conduct online from regulating access itself.

The fourth line is the moderation line of 2024. Moody v. NetChoice addressed state laws that would compel platforms to carry speech, treating a platform’s feed curation as expressive activity protected by the First Amendment and remanding for a complete facial analysis. Murthy v. Missouri addressed federal pressure on platforms to remove speech, resolving the dispute on standing without reaching the coercion merits. Together they map direct compulsion and indirect pressure from opposite directions.

The fifth line is the absent line: section 230 of the Communications Decency Act, the most litigated provision in internet law, which no opinion of the Court has ever construed. Gonzalez v. Google presented the question directly in 2023, and the Court answered the companion case instead, leaving the lower court consensus in place by default.

The account that follows traces each line in turn. The through line is the One Test answer stated above: in American internet law, constitutional doctrine carries the load that statutes were expected to bear, and the most litigated provision of all awaits its first authoritative reading.

The Foundation: Reno v. American Civil Liberties Union

Every later doctrine in internet law rests on a single holding handed down on June 26, 1997. In Reno v. American Civil Liberties Union, 521 U.S. 844 (1997), the Supreme Court struck down the online indecency provisions of the Communications Decency Act and, in doing so, fixed the constitutional status of the network itself. Speech carried over the internet would receive the full protection of the First Amendment, the protection given to print, and not the diminished protection the Court had tolerated for broadcast radio and television. That comparison, print rather than broadcast, became the foundation on which the rest of this article builds.

The Indecency Provisions in the Telecommunications Act of 1996

The provisions at issue arrived as Title V of the Telecommunications Act of 1996, Public Law 104-104, the statute that carried the indecency provisions. President Bill Clinton signed the Act on February 8, 1996, and its fifth title bore the name Communications Decency Act of 1996. The indecency measures were codified in 47 U.S.C. 223, a section of the Communications Act that had previously governed obscene telephone communications such as commercial dial-a-porn services.

Two provisions drew the constitutional challenge. The first, codified at 47 U.S.C. 223(a)(1)(B), made it a crime, by means of a telecommunications device, knowingly to make, create, solicit, or initiate the transmission of any comment, request, suggestion, proposal, image, or other communication that was obscene or indecent, knowing that the recipient was under eighteen years of age. Criminal penalties ran to imprisonment of not more than two years, a fine, or both. The second, codified at 47 U.S.C. 223(d), targeted the use of an interactive computer service. It criminalized knowingly sending to a specific person or persons under eighteen any message that, in context, depicted or described, in terms patently offensive as measured by contemporary community standards, sexual or excretory activities or organs. It also criminalized the display of such material in any manner available to a person under eighteen. The Act supplied affirmative defenses: a speaker could escape liability by taking good faith, reasonable, effective, and appropriate actions to restrict access by minors, under 223(e)(5)(A), or by requiring designated forms of age proof, such as a verified credit card or an adult identification number, under 223(e)(5)(B).

The statutory architecture contained an important internal line. Section 223(a) applied to a telecommunications device, and 223(h) specified that the term did not include the use of an interactive computer service. Section 223(d), by contrast, applied to interactive computer services expressly. Congress thus wrote one provision for one-to-one transmission and another for the multi-user services that provided access to the network. Both drew criminal sanctions, and both became the subject of a lawsuit filed the very day the Act became law.

The Three-Judge Court in Philadelphia

Section 561 of the Communications Decency Act contained a special review provision: any facial constitutional challenge to the title was to be heard by a district court of three judges convened under 28 U.S.C. 2284, with a direct appeal to the Supreme Court. On February 8, 1996, immediately after the President signed the statute, twenty plaintiffs led by the American Civil Liberties Union filed suit against Attorney General Janet Reno and the Department of Justice in the United States District Court for the Eastern District of Pennsylvania. A second suit by the American Library Association and twenty-six other plaintiffs followed. Chief Judge Dolores K. Sloviter of the Third Circuit convened the three-judge court required by the Act, and the panel consisted of Chief Judge Sloviter, District Judge Ronald L. Buckwalter, and District Judge Stewart Dalzell. The two actions were consolidated.

Within a week of filing, Judge Buckwalter entered a temporary restraining order against enforcement of section 223(a)(1)(B) insofar as it applied to indecent communications, concluding that the term indecent was too vague to support a criminal prosecution. The court then held an evidentiary hearing spanning several days, with testimony from technologists, service providers, and content publishers. On June 12, 1996, the panel issued its decision, reported at 929 F. Supp. 824, granting a preliminary injunction against enforcement of both challenged provisions. Each of the three judges wrote separately, but the judgment was unanimous.

What gave the decision its lasting power was its factual record. The three opinions together set down an extensive description of the internet as it existed in 1996, much of it based on a detailed stipulation prepared by the parties. The court found that the network was inherently decentralized, with no central control point through which content could be screened, and that a substantial share of available material, perhaps forty percent or more, originated outside the United States. It found that internet publishing worked in reverse of the broadcast model: publishers did not send material to audiences so much as make it available to readers who had to seek it out. Material did not invade the home unbidden; a user had to take affirmative steps, typing an address, following a link, or issuing a search, before anything appeared on a screen. The court distinguished among the modes of communication the network carried, including one-to-one messaging such as electronic mail, one-to-many distribution such as listservs, distributed message databases such as Usenet newsgroups, real-time communications such as Internet relay chat, remote computer utilization such as telnet, and remote information retrieval such as the World Wide Web and file transfer protocol.

The court also found that the Government’s proposed compliance mechanisms were unworkable for most speakers. Age verification by credit card or adult identification number was already in use by some commercial providers of sexually explicit material, but the court found it was not economically feasible for noncommercial speakers, who published the overwhelming share of expression on the network. The idea of tagging indecent material so that filtering software could screen it out had no established technical standard behind it. Chief Judge Sloviter concluded that the affirmative defenses were not technologically or economically feasible for most providers, and she rejected the Government’s suggestion that the statute could be narrowed by construction to reach only commercial pornographers. Judge Buckwalter held the operative terms vague under both the First and Fifth Amendments, noting that unlike obscenity, indecency had never been defined to exclude works of serious literary, artistic, political, or scientific value. Judge Dalzell, who found the statute overbroad though not vague, praised the network as a powerful engine for democratic discourse and concluded that the Government could not regulate protected expression there. The three judges disagreed on vagueness but agreed that the provisions could not be enforced.

The Supreme Court’s Holding: Strict Scrutiny and the Print Analogy

The Supreme Court heard argument on March 19, 1997, and decided the case on June 26, 1997. Justice John Paul Stevens delivered the opinion of the Court, joined by Justices Antonin Scalia, Anthony Kennedy, David Souter, Clarence Thomas, Ruth Bader Ginsburg, and Stephen Breyer. Justice Sandra Day O’Connor, joined by Chief Justice William Rehnquist, filed an opinion concurring in the judgment in part and dissenting in part. All nine Justices agreed that the Communications Decency Act’s indecency provisions could not stand in full; the division concerned how much of the statute had to fall.

Justice Stevens opened by accepting the legitimacy of the congressional goal. Protecting children from harmful materials was, in the Court’s words, a legitimate and important objective, and no one disputed it. But the indecency provisions were content-based restrictions on expression, and content-based restrictions on protected speech trigger strict scrutiny. The Government had to show not only a compelling interest but also that the statute was narrowly tailored to serve it. The Court held that the provisions failed that test twice over, as overbroad and as vague.

Vagueness came first in the analysis because it infected everything else. The terms at the heart of the statute, indecent and patently offensive as measured by contemporary community standards, gave speakers no reliable way to know what the law forbade. The Court emphasized that indecency, unlike obscenity, had never been given a limiting definition that excluded works of serious literary, artistic, political, or scientific value. A parent discussing birth control with a teenager, a librarian posting health information, or a publisher of classic literature could not determine whether a prosecutor somewhere might deem the material patently offensive. The addition of the phrase in context did not cure the problem, because the relevant context, the nature of the communication, the time of day, the medium, the identity of the speaker, might mean anything. Where a criminal statute touched expression, the Constitution demanded fair notice, and these provisions did not supply it.

Overbreadth supplied the second and more consequential ground. The provisions swept in a vast range of protected expression exchanged among adults. The Court credited the district court’s finding that most speakers on the network were noncommercial, and that the affirmative defenses were not realistic options for them. A nonprofit organization, an individual blogger, or a public library could not afford to install credit-card verification systems or to confirm the age of every reader. Faced with criminal penalties of up to two years’ imprisonment, such speakers would simply stay silent. The statute therefore operated as what the Court called a heckler’s veto by uncertainty: it suppressed far more protected expression than the interest in shielding minors could justify. As Justice Stevens put it, the interest in protecting children from potentially harmful materials did not justify an unnecessarily broad suppression of speech addressed to adults.

The heart of the opinion, and the passage on which internet law has built ever since, was the refusal to treat the network like broadcast. The Government had urged the Court to apply the reduced-scrutiny framework of FCC v. Pacifica Foundation, 438 U.S. 726 (1978), which had sustained restrictions on indecent broadcast programming. Justice Stevens rejected the invitation. Pacifica rested on three characteristics of broadcasting: a history of extensive government regulation, the scarcity of available frequencies, and the uniquely pervasive and invasive character of radio and television, which entered the home uninvited and confronted unwilling listeners and viewers. None of those characteristics fit the internet. The network had never been subject to the kind of pervasive licensing regime that governed broadcasters. Its capacity was not scarce in the broadcast sense; anyone with access could publish. And the district court’s factual findings established that internet communications were not invasive. A user had to take affirmative, deliberate steps to retrieve content, a fact that made the medium closer to a library shelf than to a radio dial. Warnings could precede material, and accidental encounters were far less likely than with broadcasting. The special justifications for diminished broadcast protection simply did not apply.

Having rejected the broadcast analogy, the Court chose the opposite pole. The internet was entitled to the full First Amendment protection given to the print press. Justice Stevens described the network as a unique medium, known to its users as cyberspace, located in no particular geographical location but available to anyone, anywhere in the world, with access. He celebrated it as a vast democratic forum, one in which any person could become a town crier with a voice that resonated farther than it could from any soapbox. That metaphor mattered doctrinally: it placed the network in the lineage of the pamphleteer and the newspaper rather than the licensed broadcaster. Strict scrutiny would apply to content-based regulation, and the Government would have to meet it.

The opinion then showed why the statute failed even on the Government’s own terms. The Government argued that less restrictive alternatives were unavailable, but the record showed otherwise. Filtering and blocking software, though imperfect, already gave parents tools to control what their children encountered, and such private controls burdened far less protected expression than a criminal ban. The Court also noted that the Government had not shown that the unregulated availability of indecent material was driving people away from the network; to the contrary, the medium was expanding dramatically. On severability, the Court accepted the Government’s argument only in part. Because obscene expression could be banned entirely, and because the prohibition on obscenity had a textual manifestation separate from the prohibition on indecency, the Court severed the words or indecent from section 223(a), leaving the rest of that subsection standing. Section 223(d) was struck in its entirety. The district court’s judgment, 929 F. Supp. 824, was affirmed.

The O’Connor Concurrence in Part and Dissent in Part

Justice O’Connor, joined by Chief Justice Rehnquist, agreed with much of the majority’s framework but parted company on how much of the statute had to fall. Her opinion treated the Communications Decency Act as, at bottom, an attempt by Congress to create adult zones on the internet, segregating indecent material into areas that minors could not enter. In her view, precedent supported the creation of such zones as constitutionally sound. She pointed to Ginsberg v. New York, 368 U.S. 629 (1968), which had sustained a prohibition on selling sexually explicit material deemed harmful to minors, and to Sable Communications of California, Inc. v. FCC, 492 U.S. 115 (1989), which had held that sexual expression which was indecent but not obscene remained protected for adults while acknowledging the Government’s interest in shielding children.

Where the majority saw provisions that inevitably chilled adult expression, Justice O’Connor saw a drafting failure. Portions of the statute were unconstitutional, she wrote, because they strayed from the blueprint the Court’s earlier cases had supplied for constructing a zoning law that could pass muster. The display provision, section 223(d)(1)(B), criminalized showing patently offensive material in any manner available to a minor, and on her reading it reached so broadly that it would effectively deny adults access to protected expression rather than merely channel it. She concurred in the judgment invalidating that provision. But she dissented from the invalidation of the narrower provisions: the indecency transmission provision, section 223(a)(1)(B), which required knowledge that the recipient was under eighteen, and the specific-person provision, section 223(d)(1)(A), which targeted the knowing sending of patently offensive material to a specific minor. Those provisions, in her view, fit the zoning blueprint. They punished identifiable knowing transmissions to children without reaching the open display of material to the public at large.

The disagreement revealed a fault line that would recur in internet law. The majority treated the network’s architecture as the decisive fact: because speakers could not reliably identify or exclude minors from open postings, any ban on indecent material available to minors would necessarily become a ban on such material available to adults. Justice O’Connor treated speaker intent as the decisive fact: a knowing transmission to a known minor was regulable even on a medium where open display was not. Her zoning theory would resurface in later litigation over successor statutes, and her partial dissent supplied the conceptual vocabulary, adult zones, cyber-zoning, for that debate. But the majority’s answer controlled: the statute as written could not separate the two audiences, and the Constitution would not permit the Government to reduce the adult population to reading only what was fit for children.

Why Reno Became the Foundation for Later Internet Speech Doctrine

Reno v. American Civil Liberties Union did more than invalidate two statutory provisions. It established the default constitutional posture of the network, and every later controversy in internet law has been argued in its shadow. The decision supplied three linked propositions that later courts have treated as settled starting points.

First, the opinion fixed the level of protection. By holding that expression on the internet receives the full First Amendment protection given to print, the Court foreclosed the argument that the medium’s novelty or its accessibility to minors justified a lesser standard. Later litigants who sought to regulate online expression had to satisfy strict scrutiny for content-based restrictions, the most demanding test in constitutional law. The Government’s interest in protecting minors, though repeatedly acknowledged as compelling, was never enough by itself; narrow tailoring and the availability of less restrictive alternatives became the decisive questions. That structure governed the litigation over the Child Online Protection Act, Congress’s successor to the Communications Decency Act, which the Court considered twice, in Ashcroft v. American Civil Liberties Union, 535 U.S. 564 (2002), and Ashcroft v. American Civil Liberties Union, 542 U.S. 656 (2004). In the second of those decisions, the Court sustained a preliminary injunction against the successor statute precisely because filtering software offered a less restrictive alternative, a direct application of Reno’s reasoning. The framework also shaped United States v. American Library Association, 539 U.S. 194 (2003), where the Court upheld the Children’s Internet Protection Act’s conditioning of federal funds on the use of filtering software in public libraries, distinguishing a funding condition from the criminal prohibition at issue in Reno. Each case negotiated within boundaries Reno had drawn.

Second, the opinion made factual findings about the medium do constitutional work. The three-judge court’s description of a decentralized, non-invasive network in which users sought out content was not background color; it was the premise of the legal analysis. The refusal to apply Pacifica turned on those findings, and the overbreadth holding turned on the finding that age verification was infeasible for noncommercial speakers. This taught later courts to begin internet speech cases with the technology as it actually operated. When the Supreme Court struck down a North Carolina statute barring registered sex offenders from social media in Packingham v. North Carolina, 582 U.S. 98 (2017), Justice Kennedy’s majority opinion described the network as the most important place for the exchange of views, a characterization that descends directly from Justice Stevens’s town crier metaphor. The doctrinal move was the same: describe the medium accurately, then assign the level of protection the description demanded.

Third, the decision drew the line between obscenity and indecency with unusual clarity and held it. Obscenity, as defined in Miller v. California, 413 U.S. 15 (1973), and child pornography, as defined in New York v. Ferber, 458 U.S. 747 (1982), remained wholly unprotected, and the Court preserved the Government’s authority to prosecute both by severing or indecent from section 223(a) rather than striking the subsection whole. But indecent expression that fell short of obscenity, sexual expression that was merely patently offensive, retained full protection for adults. That distinction organized the next decade of legislation. Congress could criminalize obscenity and the sexual exploitation of children online without constitutional difficulty; when it reached for the broader category of material harmful to minors, it had to reckon with Reno.

The opinion’s influence extended beyond the courtroom. The district court’s factual findings became the standard judicial description of the early internet, quoted in briefs, law review articles, and subsequent opinions for years. The decision also vindicated a strategic choice by the challengers: by assembling a plaintiff coalition that included libraries, publishers, civil liberties organizations, and ordinary speakers, and by building a meticulous evidentiary record about how the network functioned, they gave the courts a factual foundation that made the constitutional conclusion difficult to resist. Doctrine followed description.

Reno’s limits deserve acknowledgment as well. The decision addressed only the indecency provisions before it; it did not immunize the network from all regulation, and it left open how the First Amendment would apply to conduct-facilitating uses of the medium, to commercial fraud, or to the Government’s own funding conditions. Justice O’Connor’s partial dissent preserved the argument that narrower, intent-focused provisions could survive, an argument Congress tested with later statutes. But as a foundation, the decision has held. The holding that the internet receives the full protection of the First Amendment, rather than the diminished protection of broadcast, remains the premise from which all later internet speech doctrine proceeds. Everything else in this article’s account of internet law, the successor statutes, the filtering mandates, the platform disputes, and the social media cases, is an elaboration of, a qualification of, or an argument within the framework that Reno established in 1997.

What two funding streams did CIPA condition?

CIPA conditioned two federal funding streams for Internet access: discounts under the E-rate program, created by the Telecommunications Act of 1996 to subsidize connectivity for schools and libraries, and grants under the Library Services and Technology Act. A library that declined the filtering condition could forgo both and operate unfiltered.

What did the American Library Association plurality compare library filtering to?

The plurality compared filtering to the collection decisions libraries have always made. A library acquires Internet access to serve research, learning, and recreation, the opinion reasoned, much as it acquires books, and selecting what the terminals would carry was the digital equivalent of a librarian choosing what belongs on the shelves.

Why did Kennedy’s CIPA concurrence make administration decisive?

Justice Kennedy joined the judgment sustaining CIPA only because the statute let libraries disable filtering for adult patrons without significant delay. On his account, the scheme’s constitutionality depended on its administration: a library that refused to disable blocking, or burdened such requests, could lose an as-applied challenge.

What did the first COPA decision actually settle?

The 2002 decision settled one point only: COPA’s reliance on contemporary community standards, standing alone, did not render the statute substantially overbroad. The Court expressed no view on vagueness, on other overbreadth theories, or on strict scrutiny, and the preliminary injunction against enforcement remained in place.

What did the Court hold in Murthy v. Missouri?

In Murthy v. Missouri, 603 U.S. 43 (2024), the Court held that the plaintiffs had failed to establish standing to challenge federal officials’ communications with platforms. Because standing failed, the Court never applied the Bantam Books coercion test to the officials’ conduct, and the governing framework from 1963 remained in place unapplied.

The Funding Condition Line

In the years after Reno v. American Civil Liberties Union, 521 U.S. 844 (1997), struck down the Communications Decency Act, Congress returned to the problem of shielding children from sexual material on the Internet with two different instruments. One approach attached strings to federal money. The other imposed criminal penalties directly on speakers. The Supreme Court treated those two approaches as fundamentally different exercises of congressional power, and the contrast between them became one of the most durable lines in internet law.

The funding instrument was the Children’s Internet Protection Act of 2000 (CIPA), and it survived in United States v. American Library Association, 539 U.S. 194 (2003). The criminal instrument was the Child Online Protection Act of 1998 (COPA), and it never survived its preliminary injunction, losing first in Ashcroft v. American Civil Liberties Union, 535 U.S. 564 (2002), and again in Ashcroft v. American Civil Liberties Union, 542 U.S. 656 (2004), before a permanent injunction and the denial of certiorari ended the statute in 2009. The difference between the two outcomes turned on an old constitutional principle: Congress may attach conditions to money it need not spend, while direct prohibitions on protected expression face the strictest judicial scrutiny.

The pairing offers a controlled comparison that students of internet law return to often. Both statutes pursued the same compelling interest, protecting minors from sexual material online. Both were enacted within two years of each other. Yet one was framed as Congress deciding how to spend its own money, and the other as Congress deciding what speakers could say under threat of imprisonment. The Supreme Court’s divergent treatment of the two measures confirmed that the form of the regulation, not only its purpose, determines the constitutional analysis.

Congress’s Money, Congress’s Conditions

CIPA was enacted in 2000 as part of a large appropriations measure. It addressed two federal funding streams that helped libraries and schools pay for Internet access: discounts under the E-rate program, created by the Telecommunications Act of 1996 to subsidize telecommunications and connectivity for schools and libraries, and grants under the Library Services and Technology Act. CIPA provided that a public library could not receive those funds for Internet access unless it installed software to block visual depictions that constituted obscenity or child pornography and, on terminals accessible to minors, material defined as harmful to minors. The statute permitted libraries to disable the blocking for adult patrons to allow access for bona fide research or other lawful purposes, and to unblock individual sites on request.

The American Library Association, joined by public libraries, library patrons, and publishers of Web sites, challenged the statute. A three-judge panel of the United States District Court for the Eastern District of Pennsylvania held the provisions facially invalid in 2002, reasoning in 201 F. Supp. 2d 401 that the filtering mandate was a content-based restriction on expression in a designated public forum, subject to strict scrutiny, and not narrowly tailored to the government’s interest. The district court found that available filtering software inevitably blocked large quantities of constitutionally protected material along with the unprotected, and that the statute’s disabling and unblocking provisions were an inadequate cure, since some patrons would be embarrassed to ask for access and some libraries could not process requests promptly. The case came to the Supreme Court on direct appeal, was argued on March 5, 2003, and was decided on June 23, 2003.

Congress had written an expedited review provision into the statute, and the challenge therefore traveled to the Supreme Court on direct appeal from the three-judge district court rather than through the ordinary court of appeals route. The plaintiffs spanned the library world and the publishing world: the American Library Association, individual public libraries and their patrons, and Web publishers whose sites had been blocked by filtering programs. Their facial attack asked the Justices to invalidate CIPA in all its applications, the most demanding form of constitutional challenge.

A Plurality Sustains the Condition

Six Justices voted to reverse and sustain CIPA against the facial challenge, but no single rationale commanded five votes. Chief Justice Rehnquist announced the judgment and delivered a plurality opinion joined by Justices O’Connor, Scalia, and Thomas. Justice Kennedy and Justice Breyer each filed opinions concurring in the judgment. Justice Stevens dissented, and Justice Souter dissented in an opinion joined by Justice Ginsburg.

The plurality’s central move was to treat CIPA as an exercise of the spending power rather than as a regulation of speech. Congress was not required to subsidize Internet access in public libraries at all, the plurality reasoned, and when Congress appropriates public funds to establish a program it may define the limits of the program it chooses to fund. The opinion drew on South Dakota v. Dole, 483 U.S. 206 (1987), and Rust v. Sullivan, 500 U.S. 173 (1991), for the proposition that a refusal to fund protected activity is not the same as a penalty on it. Because CIPA imposed no sanction on any speaker and left every library free to decline federal assistance and operate unfiltered terminals, the plurality concluded that the statute did not impose an unconstitutional condition on the receipt of federal funds.

The plurality also rejected the public forum analysis that had driven the district court’s strict scrutiny holding. Internet terminals in public libraries, the plurality held, were neither traditional public forums nor designated ones. A library acquires Internet access to facilitate research, learning, and recreational pursuits, the opinion explained, much as it acquires books, and decisions about what material to make available through those terminals were analogous to the collection decisions libraries have always made. Filtering software, on this view, was a technological means of carrying out the same kind of selection that librarians perform with shelves.

The plurality acknowledged the evidence that filtering programs block protected expression by mistake, but it treated the statute’s disabling provisions and the government’s representations as the answer. At oral argument the Solicitor General had represented that libraries could readily disable blocking for adult patrons and unblock individual sites, and the plurality credited that representation. If experience proved otherwise, the plurality added, libraries and patrons could bring as-applied challenges; the facial attack before the Justices did not require the statute’s invalidation.

The forum analysis drew a sharp line between a library’s shelves and a public square. The plurality stressed that a public library does not install Internet terminals to create a forum for Web publishers to address the public; it installs them so that patrons can pursue research, learning, and recreation. Because the terminals existed to serve the library’s own expressive and educational mission, decisions about blocking were treated as the digital equivalent of a librarian’s choice about which materials to acquire. The First Amendment, the plurality concluded, does not compel a library to furnish access to pornography simply because it has chosen to furnish access to the Internet.

Justice Kennedy concurred in the judgment on narrower ground. He agreed that CIPA could be sustained, but only because the statute permitted libraries to disable filtering for adults without significant delay. The constitutionality of the scheme, in his account, depended on how libraries actually administered it: a library that refused to disable blocking for an adult patron, or that imposed burdensome procedures on such requests, could face a successful as-applied challenge. The saving construction was not decorative. It was the condition on which the judgment rested.

Justice Breyer also concurred in the judgment, applying a form of heightened scrutiny that balanced the statute’s expression-related burdens against its objectives. The burden on adult library users was minimal, he concluded, precisely because the blocking could be lifted easily on request, a point on which he too relied on the Solicitor General’s representations. The government’s interest in preventing the federal subsidy from underwriting access to pornography in public institutions justified that modest burden.

Justice Stevens dissented. In his view CIPA coerced public libraries into acting as censors, imposing a content-based restriction on patrons’ access to protected expression. Children could be shielded by less restrictive means, he argued, such as filtering only the terminals designated for minors’ use, and the disabling provisions placed the burden on adults to identify themselves and ask permission to view lawful material, a requirement that would chill the exercise of First Amendment rights. The plurality’s dependence on oral representations by government counsel, he suggested, was a thin foundation on which to rest the rights of library patrons.

Justice Souter, joined by Justice Ginsburg, dissented on related but distinct grounds. He treated the library’s Internet terminals as a designated public forum for the receipt of information, making the filtering mandate a content-based exclusion subject to strict scrutiny. The statute’s condition, he argued, was unconstitutional precisely because it required libraries to surrender First Amendment protections as the price of federal money, and the disabling mechanism did not cure the violation: it forced patrons seeking protected material to request special treatment, deterring the very access the First Amendment secures.

Conditions Versus Prohibitions

The doctrine that carried CIPA has deep roots. The unconstitutional conditions principle holds that the government may not condition the receipt of a benefit on the surrender of a constitutional right, yet the Supreme Court has long distinguished between conditions that define the boundaries of a funded program and conditions that penalize protected activity beyond it. South Dakota v. Dole sustained conditions on highway funds because they specified how federal money would be used. Rust v. Sullivan sustained restrictions on abortion counseling within a federally funded family planning program because the restrictions defined the program Congress had chosen to subsidize, leaving recipients free to engage in the restricted activity with other funds. Legal Services Corporation v. Velazquez, 531 U.S. 533 (2001), went the other way, striking down a condition that distorted the functioning of the courts themselves.

CIPA fell on the permissible side of that line because it regulated only the federally subsidized activity and imposed no consequence beyond the loss of the subsidy. A library that found the filtering requirement unacceptable could simply forgo E-rate discounts and technology grants. The contrast with Reno was decisive. The Communications Decency Act had imposed criminal penalties on speakers themselves, reaching every person who transmitted indecent material where minors might encounter it, regardless of any federal funding. CIPA punished no one. It offered money with strings, and strings on money the government need not offer are judged under the spending power, not under the strict scrutiny reserved for direct prohibitions of protected expression. That distinction is the core of the funding condition doctrine, and it explains why two statutes aimed at the same social problem met opposite fates.

The fit between CIPA’s condition and the funded program also mattered. In Dole the Justices had sustained highway fund conditions because they served the federal interest in safe interstate travel; a condition related to the purpose of the expenditure stands on firmer ground than one that reaches beyond it. CIPA’s filtering requirement was tied directly to the federal dollars at issue: Congress was paying for Internet access in libraries and schools, and it declined to pay for access to pornographic images on those subsidized terminals. The condition was also noncoercive in the Dole sense. The federal contribution was a discount and a grant, not the lifeblood of library budgets, and any institution that objected could decline the money and operate without filters. Velazquez marked the boundary on the other side: there the condition had distorted an independent system, the courts, by forbidding funded lawyers to make arguments the adversary process required. CIPA distorted nothing; it defined what the subsidy would buy.

The Injunction COPA Never Escaped

Congress had enacted the Child Online Protection Act in October 1998, and President Clinton signed it, as a direct response to the invalidation of the Communications Decency Act in Reno. Where the earlier statute had criminalized indecent communications, COPA narrowed its reach to commercial Web communications containing material harmful to minors, tracking the obscenity-related definitions of Miller v. California, 413 U.S. 15 (1973), and Ginsberg v. New York, 390 U.S. 629 (1968). The statute imposed fines of up to 50,000 dollars per day and imprisonment of up to six months on those who knowingly made such material available to minors for commercial purposes. It offered affirmative defenses to operators who restricted access through credit card verification, adult access codes, digital certificates, or other reasonable measures feasible under available technology.

The statute never took effect. In February 1999, before any prosecution could be brought, the Eastern District of Pennsylvania granted a preliminary injunction against COPA’s enforcement in a suit brought by the American Civil Liberties Union and a group of Web publishers that included an online magazine, a sexual health information site, a gay newspaper, and the creator of an online slang dictionary. The district court, in 31 F. Supp. 2d 473, found the plaintiffs likely to succeed on their First Amendment claim. The Third Circuit affirmed in 2000, 217 F.3d 162, but on a ground the district court had not relied upon: the court of appeals held that COPA’s use of contemporary community standards to identify material harmful to minors rendered the statute substantially overbroad.

The Supreme Court first addressed COPA in Ashcroft v. American Civil Liberties Union, 535 U.S. 564, decided May 13, 2002. Justice Thomas announced the judgment in an opinion joined in full by only some of his colleagues, with other portions commanding different alignments among the participating Justices. Justices O’Connor and Breyer each filed opinions concurring in part and concurring in the judgment, Justice Kennedy filed an opinion concurring in the judgment joined by Justices Souter and Ginsburg, and Justice Stevens dissented.

The fragmentation mattered. Justice Kennedy’s opinion concurring in the judgment, joined by Justices Souter and Ginsburg, agreed that the case should return to the lower courts but warned that the community standards problem might still prove fatal and that COPA could fail strict scrutiny on remand. Justice Stevens, dissenting alone, would have gone further and held that reliance on community standards did render the statute substantially overbroad. The result was a decision that resolved almost nothing about COPA’s constitutionality while leaving the injunction firmly in place.

The 2002 holding was narrow. The Justices agreed that COPA’s reliance on contemporary community standards, standing alone, did not render the statute substantially overbroad for First Amendment purposes. The opinion expressed no view on whether COPA suffered from overbreadth for other reasons, whether it was unconstitutionally vague, or whether it could survive strict scrutiny, leaving those questions for the court of appeals on remand. Because the government had not asked the Supreme Court to vacate the preliminary injunction, and because vacating it would have required deciding issues the Third Circuit had not yet addressed, the injunction remained in place. COPA stayed unenforced.

On remand the Third Circuit again affirmed the preliminary injunction, this time in 322 F.3d 240 (2003), holding that COPA was not the least restrictive means of achieving the government’s compelling interest in protecting minors. Filtering software installed by parents and schools, the court concluded, was a less restrictive and at least as effective alternative to criminal penalties on publishers. The government sought review once more.

The second Supreme Court encounter, Ashcroft v. American Civil Liberties Union, 542 U.S. 656, was decided June 29, 2004. Justice Kennedy delivered the opinion of the Court, joined by Justices Stevens, Souter, Thomas, and Ginsburg. Justice Stevens filed a concurring opinion joined by Justice Ginsburg. Justice Scalia dissented alone, and Justice Breyer dissented in an opinion joined by Chief Justice Rehnquist and Justice O’Connor.

The 2004 decision upheld the preliminary injunction. Applying the abuse of discretion standard that governs review of such orders, the majority held that the district court had not erred in finding the plaintiffs likely to prevail. The government, the opinion explained, had failed to rebut the plaintiffs’ contention that plausible less restrictive alternatives to COPA existed and would be at least as effective in shielding minors from harmful material. Filtering software, already credited by the lower courts, was the centerpiece of that showing. The majority also noted that the legal landscape had shifted since the district court’s findings, with Congress having enacted additional measures, such as a prohibition on misleading domain names and a child-safe domain, that might themselves qualify as less restrictive alternatives. The case was remanded for trial on the merits, and the preliminary injunction against COPA’s enforcement continued.

The dissents gave the opposing account of filtering. Justice Breyer, joined by Chief Justice Rehnquist and Justice O’Connor, argued that Congress could reasonably have concluded that a regime built entirely on filtering software was not an effective system: filters underblocked, imposed costs on each family that used them, failed to screen children’s access outside the home, and lacked precision in distinguishing protected from unprotected expression. On that view COPA’s age verification screen was a modest burden on adults that complemented filters rather than duplicating them, and the statute deserved to survive strict scrutiny. Justice Scalia dissented separately and briefly, taking the position that COPA was a constitutional regulation of commercial pornography.

A Statute Enjoined Before It Ever Took Effect

The trial the 2004 decision ordered produced the statute’s final defeat. After a full bench trial, the Eastern District of Pennsylvania, in an opinion by Judge Reed, held COPA facially unconstitutional and converted the preliminary restraint into a permanent injunction on March 22, 2007, in ACLU v. Gonzales, 478 F. Supp. 2d 775. Applying strict scrutiny to the content-based criminal prohibition, the district court concluded that the statute was not narrowly tailored to the government’s compelling interest and that less restrictive alternatives, particularly filtering software promoted and improved since the litigation began, were available and effective.

Preparing for that trial, the government had sought evidence that filtering software was less effective than COPA’s screening requirement. In 2005 the Department of Justice subpoenaed several search engine companies for large samples of search records. Google resisted, arguing that compliance would burden its operations and shake public confidence in the privacy of search. In March 2006 Judge James Ware of the Northern District of California narrowed the demand, ordering Google to produce a sample of 50,000 indexed Web addresses while refusing the government’s request for 5,000 search queries. America Online, Microsoft’s MSN, and Yahoo had already supplied some data. The episode became one of the earliest public confrontations over government access to search engine records, and it grew directly out of the constitutional test the Supreme Court had imposed: the government had to prove, with evidence, that no less restrictive alternative would serve as well.

The Third Circuit affirmed on July 22, 2008, in a unanimous panel decision, 534 F.3d 181, agreeing that COPA was impermissibly overbroad and vague and that filters offered a less restrictive path to the statute’s goals. The government sought certiorari a third time. On January 21, 2009, the Supreme Court declined review, leaving the permanent injunction undisturbed. COPA, enacted more than a decade earlier, had never been enforced against a single speaker.

The two statutes thus trace a clean arc through internet law. CIPA endured because it was a condition on a subsidy: Congress offered money for Internet access in libraries and defined the terms of its own offer, leaving institutions free to walk away. COPA perished because it was a criminal prohibition aimed directly at publishers: it threatened fines and imprisonment for protected expression, and under strict scrutiny the existence of filtering software proved fatal not once but repeatedly. The same technology that rescued the funding condition, by making the disabling of library filters a practical reality, doomed the criminal statute, by giving judges a less restrictive alternative to punishment. How Congress chose to legislate mattered as much as what it chose to legislate.

The Access Line: Packingham v. North Carolina

Packingham v. North Carolina, 582 U.S. 98 (2017), decided on June 19, 2017, asked whether a state could make social media off-limits to an entire class of people. The Supreme Court answered unanimously in the judgment, with Justice Gorsuch not participating, that North Carolina could not. Justice Anthony Kennedy wrote the majority opinion for himself and Justices Ginsburg, Breyer, Sotomayor, and Kagan, holding that the challenged statute violated the First Amendment. Justice Alito, joined by Chief Justice Roberts and Justice Thomas, concurred in the judgment only, agreeing that the particular law had to fall but warning that the majority’s broad descriptions of cyberspace might be misread in later disputes. The decision became a landmark in internet law, the first Supreme Court case to treat access to social networking platforms as constitutionally significant in its own right, while carefully reserving questions it did not answer.

Before Packingham, the Supreme Court’s principal statement on speech and the internet remained Reno v. ACLU, 521 U.S. 844 (1997), which had struck down federal indecency restrictions on online communication. Two decades separated the two decisions, and the distance measured more than time. Reno confronted a medium still taking shape; Packingham confronted a settled social infrastructure, one in which hundreds of millions of Americans conducted their civic, professional, and religious lives. The litigation history of internet law between those bookends had produced doctrines for commercial speech, for obscenity, and for the liability of intermediaries, but no case had asked the Court to weigh the constitutional significance of access itself, of simply being present where the exchange of views happened.

The section’s title draws on the distinction the case forced: between regulating conduct on the internet and regulating access to it. North Carolina had drawn the line at the doorway, criminalizing presence. The Supreme Court redrew it around conduct, permitting states to police what offenders do online while protecting their ability to be online at all. That distinction has since organized much of the doctrine governing digital restrictions on disfavored speakers.

The North Carolina statute

The law at issue was N.C. Gen. Stat. section 14-202.5, enacted in 2008 as part of the legislature’s effort to address online risks to minors. It made it a felony for a person registered as a sex offender under Article 27A of Chapter 14 to access a commercial social networking website, where the offender knew the site permitted minor children to become members or to create or maintain personal web pages there. The definition of a commercial social networking website, drawn from G.S. 75-40, turned on four functional criteria: the site had to be operated by someone who derived revenue from membership fees, advertising, or related sources; it had to facilitate the social introduction of two or more persons; it had to allow users to create pages or profiles containing names or nicknames, photographs, and other personal information; and it had to provide users or visitors with a mechanism for communicating with one another, such as a message board, chat room, or instant messenger.

The measure emerged from the 2007 session of the North Carolina General Assembly as Senate Bill 132, a broader package of legislation addressing sexual offenses, and took effect in 2008. Its premise was straightforward: social networking services had become venues where adults could misrepresent themselves, cultivate contact with minors, and gather personal information about children. Rather than prohibiting those predatory acts online, however, the legislature prohibited a class of people from entering the venues altogether. The offense required knowledge: a registrant violated the law only where he knew the site permitted minors to become members or to create personal web pages. Violation was punishable as a felony, and jurisdiction attached where the transmission either originated in the state or was received there.

The breadth of that definition was the law’s central vulnerability. As the litigation later established, the four criteria swept far beyond the recognizable social networking services the legislature had in view. News sites with comment sections, online retailers with user accounts and review forums, and health information portals all plausibly satisfied every element: they generated revenue, enabled introductions or exchanges among strangers, hosted user profiles, and supplied communication mechanisms. A registrant seeking to read a newspaper article, comparison-shop for a household purchase, or look up medical information online risked a felony charge for logging in. The statute did carve out two narrow exclusions, one for sites providing only a single discrete service such as photo sharing, email, instant messaging, or a chat room, and another for sites whose primary purpose was facilitating commercial transactions in goods or services between members or visitors. Those exclusions left the core problem intact, because a great many ordinary websites combined a profile system, a comment function, and advertising revenue without having commercial transactions or a single discrete service as their purpose. State authorities charged more than a thousand registrants under the provision before its constitutional fate was settled, with enforcement driven in several counties by officers who monitored social media specifically to identify registered offenders behind alias accounts.

A traffic ticket, a Facebook post, and a felony charge

The test case arose from facts almost absurdly far from the statute’s purpose. In April 2010, Lester Gerard Packingham, a registered sex offender, opened a Facebook account under the pseudonym J.R. Gerrard. He had beaten a traffic citation in court and posted a brief message thanking God the ticket had been dismissed. The post involved no minors, no solicitation, and no conduct bearing any connection to the sexual offenses that had placed Packingham on the registry.

A Durham police officer assigned to monitor social media for registrants discovered the account, identified Packingham as its operator, and charged him with violating section 14-202.5.

Packingham’s prosecution grew out of a deliberate enforcement program. The Durham officer who found the alias account was part of a unit assigned to monitor social networking sites for registrants, cross-referencing online identities against the state registry. The pseudonym J.R. Gerrard, a near-anagram of Packingham’s own name, made identification straightforward once the account surfaced. Nothing in the resulting charge alleged danger to a minor; the state’s theory was access itself, the mere presence of a registrant on a covered platform. The case thus presented the statute in its purest form, unclouded by any allegation of the predatory conduct the law was meant to prevent, and it forced courts to decide whether presence alone could be criminalized. Packingham moved to dismiss, arguing that the statute violated his rights under the First Amendment. The trial court denied the motion, and Packingham was convicted. The facts underscored the measure’s sweep: conduct with no demonstrated risk to anyone, expression of ordinary gratitude in a forum used by hundreds of millions, was enough to sustain a felony prosecution.

The North Carolina courts divide

The intermediate appellate court sided with Packingham. In 2013, the North Carolina Court of Appeals reversed the conviction, holding in 749 S.E.2d 842 that the statute violated the First Amendment. The court treated the measure as a content-neutral restriction on speech and applied intermediate scrutiny, accepting the state’s interest in child protection as significant while concluding the law was not narrowly tailored to serve it. The breadth of the definition, the opinion reasoned, restricted far more expression than the protective purpose required, and the state had not shown a blanket access ban was necessary where narrower alternatives existed.

That ruling kept Packingham’s conviction aside for roughly two years before the state’s highest court intervened. The Supreme Court of North Carolina reversed the Court of Appeals in 2015 and reinstated the conviction, reasoning that the measure was narrowly tailored to the significant government interest in protecting children from sexual predators. On the state high court’s reading, the law did not impose a blanket prohibition on internet use; it addressed only those websites that afforded an opportunity to gather information about minors, and a registrant remained free to use the rest of the web. The majority also stressed the practical difficulty of policing individual online contacts with minors, treating a broad access restriction as a reasonable response to an enforcement problem the state could not solve through case-by-case supervision.

The division set up the federal question cleanly. The state supreme court had treated the restriction as targeted and proportional. The Court of Appeals had treated it as a broad speech restriction that the state had failed to justify. The United States Supreme Court granted review to resolve which characterization the First Amendment supported.

Kennedy’s majority: cyberspace and the exchange of views

Justice Kennedy’s majority opinion opened with a proposition that became the decision’s most quoted holding: the case was among the first the Court had taken to address the relationship between the First Amendment and the internet, and courts should exercise extreme caution before suggesting that constitutional protection was thin for access to the vast networks of that medium. From there, the opinion built its account of what social media had become. Kennedy described cyberspace and the social networking platforms within it as among the most important places for the exchange of views. For many people, the opinion observed, those services were the principal sources for learning current events, checking advertisements for employment, speaking and listening in the public square, and otherwise exploring the vast realms of human thought and knowledge.

The reasoning proceeded in two steps. First, the Court accepted that North Carolina’s interest in preventing sexual crimes against children was legitimate and important. Second, it held that the statute was not narrowly tailored to that interest even under intermediate scrutiny, the standard for content-neutral regulations. The opinion did not dispute the gravity of the state’s objective or the difficulty of detecting predatory contact online. It held instead that a legislature confronting a genuine enforcement problem may not answer it with a restriction reaching far beyond the problem’s boundaries. The statute did not punish contacting a minor, gathering information about a minor, or misrepresenting one’s identity to a minor; it punished logging in. By framing the issue as access rather than conduct, the majority converted what North Carolina had presented as a protective regulation into a speaker-based exclusion from the principal forums of contemporary expression, and speaker-based exclusions demand a precision the statute lacked. The decisive flaw was scope. With one broad stroke, as the opinion put it, the state had barred registrants from the very channels through which so much contemporary expression, information gathering, and civic participation flowed. To foreclose access to social media altogether, Kennedy reasoned, was to prevent the user from engaging in the legitimate exercise of First Amendment rights. The opinion added that it was unsettling to suggest that only a limited set of websites could be used even by persons who had completed their sentences, and that even convicted criminals, in some instances especially convicted criminals, might receive legitimate benefits from these means of access to the world of ideas, particularly where they sought to reform and to pursue lawful and rewarding lives.

The majority emphasized that the definition’s reach confirmed the overbreadth. A law written to keep offenders from contacting minors online had been drafted so broadly that it covered sites with no plausible connection to predatory conduct: news outlets, retail platforms, and health resources all fell within the four-part test. That mismatch meant the statute restricted an enormous quantity of harmless and constitutionally valued activity while adding little to the state’s protective aims. Kennedy suggested that the state had narrower tools available, pointing to prohibitions on the conduct that often presages a sexual crime, such as contacting a minor or using a website to gather information about a minor. Those targeted measures, the Court implied, would serve the protective interest without cutting off access to the principal forums of contemporary expression. In internet law, the holding established that a complete bar on access to social media by a class of offenders could not survive First Amendment review, and it did so by treating the access restriction as a burden on the speaker’s own expressive activity rather than as a regulation of conduct alone.

The opinion also carried a forward-looking caution that later opinions in this area would echo. By stressing that courts should exercise extreme caution before suggesting scant protection for access to vast online networks, Kennedy signaled that the constitutional framework built for streets, parks, and pamphlets would have to be applied with care to a medium whose architecture was still changing. The line served both as a doctrinal guide and as an admission of uncertainty: the Court was deciding a case about the internet of 2017 while aware that later technologies might demand fresh analysis.

The Alito concurrence: a warning about sweeping language

Justice Alito, writing for himself, Chief Justice Roberts, and Justice Thomas, agreed that North Carolina’s law was unconstitutional but declined to join the majority’s reasoning. The concurrence accepted the outcome on a narrower ground: the statutory definition was so capacious that it barred registrants from using websites, such as Amazon, the Washington Post, and WebMD, that were most unlikely to facilitate the commission of a sex crime against a child. On that basis alone, the law could not stand.

The concurrence’s significance lay in its caution about the majority’s rhetoric. Alito rejected the characterization of cyberspace as the functional equivalent of public streets and parks in the physical world, arguing that there were important differences between the two environments. The majority’s sweeping statements about the internet, the concurrence warned, could be misread in future cases in ways that would leave states with little ability to restrict access to sites that might be visited by even the most dangerous sex offenders. Where Kennedy’s opinion treated the analogy between online platforms and traditional public forums as a reason to protect access, Alito insisted that the analogy was imperfect and that the Court should not commit itself to doctrinal language that later litigants would stretch beyond this dispute.

The disagreement mattered for the opinion’s precedential weight. Because three justices joined the judgment without joining the reasoning, the broadest propositions in Kennedy’s opinion, the public-square analogies and the pronouncements about cyberspace, spoke for five justices rather than eight. The concurrence signaled to lower courts and legislatures that the outcome rested on the statute’s extraordinary breadth, not on a holding that online platforms occupied the same constitutional position as physical public forums in every respect. Alito’s framing left room for a future case to sustain a carefully drawn restriction on an offender’s use of a particular site or service, provided the state could show the measure was genuinely tied to preventing sexual predation rather than to walling off the internet at large. In that sense the concurrence read Packingham as a decision about tailoring, not as a charter of platform access, and its narrower rationale has given state attorneys general and legislators a template for arguing that their own measures fall on the permissible side of the line the case drew.

The concurrence also pressed a doctrinal objection. Kennedy’s opinion had treated social media as analogous to the public square, a characterization carrying echoes of public-forum doctrine, under which exclusions from traditional public spaces face the strictest scrutiny. Alito rejected the analogy, insisting cyberspace differed from physical public spaces in ways bearing on regulation: platforms are privately owned, architected, and moderated, and the state’s relationship to them is not the relationship of a government to its own streets and parks. A court treating the analogy as settled doctrine, the concurrence warned, would hamstring legislatures attempting to regulate genuinely dangerous online spaces, including services designed around contact between strangers. The warning gave lower courts a basis for reading Packingham’s broadest language as descriptive of the statute before the Court rather than as a general theory of the internet.

What Packingham established and what it left open

Packingham established a proposition that reshaped internet law: access to social networking platforms is constitutionally significant, and a state may not impose a complete bar on such access as a condition attached to a broad class of offenders. The decision recognized that participation in the exchange of views had migrated to these services to a degree that cutting someone off from them entirely imposed a burden the First Amendment would not tolerate without a far more precise justification. That recognition has since been invoked in litigation over prison and probation restrictions on internet use, conditions of supervised release, and other measures that limited offenders’ access to online expression. Federal appellate courts in the years after 2017 cited the decision when reviewing supervised-release conditions that barred internet access outright, using its tailoring analysis to distinguish between total prohibitions, which Packingham condemned, and focused restrictions on particular uses, which it left open.

The opinion’s limits were equally deliberate. The Court did not hold that registrants possess an affirmative right of access to social media in all circumstances. It did not address whether a more precisely drawn law, one aimed at actual contact with minors, at impersonation of children online, or at the use of websites to gather information about minors, would survive scrutiny; Kennedy’s opinion suggested that such narrower measures remained available. The Court did not disturb restrictions imposed as individualized conditions of probation or supervised release, where a sentencing judge tailors limitations to a particular offender’s history and risk, nor did it speak to internet access within prisons and jails, where the constitutional analysis follows different doctrines built around the special needs of incarceration. The unanimous judgment masked the methodological split that Alito’s concurrence preserved, and that split has shaped how carefully lower courts apply Packingham’s broad language to new facts. Judges confronting prison internet policies or platform-level age restrictions have treated the case as establishing the constitutional weight of online access while testing each new restriction against the tailoring standard the majority applied.

The aftermath confirmed the decision’s practical effect. North Carolina rewrote the statute in 2019 through Session Law 2019-245, replacing the blanket prohibition on access with targeted bans on specific online conduct by high-risk offenders: communicating with a person believed to be under sixteen, contacting such a person, falsely posing as a minor with criminal intent, using a website to gather information about a minor, and using social networking sites in violation of posted policies excluding convicted offenders. The revision moved the state’s approach toward the narrower measures the Kennedy opinion had gestured toward, trading a total access bar for prohibitions tied to the predatory conduct the legislature had originally sought to prevent. In that sense Packingham did not end regulation of offenders’ online activity; it redirected that regulation from the doorway of the forum to the conduct within it. The case stands in internet law as the moment the Supreme Court acknowledged that the right to speak and to listen had followed the public into cyberspace, and that any law seeking to lock the door behind a speaker would have to justify itself against the full weight of what those networks had become.

The lineup itself carried meaning. The judgment was unanimous, eight to zero, with Justice Gorsuch taking no part, having joined the Court after argument. Unanimity on the result, paired with division on the reasoning, signaled that the statute’s defect was obvious across methodological divides while the constitutional status of the internet remained contested. Kennedy’s closing caution, that courts should hesitate before suggesting scant protection for access to vast online networks, was joined by five; Alito’s counter-caution, against reading the opinion as a license for unfettered access in every future dispute, was joined by three. Between them, the two opinions mapped the terrain on which later internet law would be argued: access is constitutionally significant, breadth is fatal, and the analogies courts choose will decide how far the principle travels.

The Moderation Line: The 2024 State-Law Decisions and Government Jawboning

Two Statutes, Two Circuits

In 2021, Texas and Florida each enacted a statute aimed at large internet platforms and their handling of user content, and the two laws took different regulatory approaches to the same concern. Texas House Bill 20 barred covered platforms from censoring expression based on the viewpoint of the speaker, creating a cause of action that allowed users to sue companies over removals and other moderation decisions tied to viewpoint. Florida Senate Bill 7072 prohibited large platforms from deplatforming political candidates, imposed fines for doing so, and required detailed disclosures of moderation policies, treating the covered services in some respects like common carriers. Both statutes reached entities defined by size thresholds, including revenue above one hundred million dollars or monthly audiences above one hundred million users, and both were challenged immediately by NetChoice, a trade association representing major platform operators, which argued that the restrictions violated the First Amendment.

The two challenges produced opposite results in the lower courts. In the Florida litigation, the Eleventh Circuit largely upheld a district court injunction against the state law, concluding that the measure burdened the editorial judgment of private companies in a way the First Amendment did not permit. In the Texas litigation, the Fifth Circuit reversed the district court’s injunction and sustained House Bill 20, reasoning that the platforms were not speaking through their moderation choices and that the state could require viewpoint-neutral carriage of user expression. Before the merits were decided, the Texas law had already generated a shadow-docket episode: after the Fifth Circuit stayed the district court’s injunction in May 2022, the Supreme Court vacated that stay on May 31, 2022, in a 5-4 order with no written reasoning from the majority, restoring the injunction and keeping House Bill 20 unenforced while the litigation continued. Justice Alito dissented from that order, joined by Justices Thomas and Gorsuch, and Justice Kagan noted her dissent separately. The resulting split between the circuits on laws with similar subject matter made Supreme Court review nearly certain, and the Court granted certiorari in both matters, hearing argument on February 26, 2024, in Moody v. NetChoice, LLC (No. 22-277) and NetChoice, LLC v. Paxton (No. 22-555).

The Kagan Majority and the Facial-Challenge Procedure

On July 1, 2024, the Court issued a single opinion covering both cases, authored by Justice Kagan and joined in full by Chief Justice Roberts and Justices Sotomayor, Kavanaugh, and Barrett, with Justice Jackson joining as to Parts I, II, and III-A. The bottom-line judgment was unanimous in form: the judgments below were vacated and both cases remanded. The basis for that disposition was procedural rather than substantive. NetChoice had brought facial challenges, asserting that each statute violated the First Amendment on its face and across all of its applications. A proper facial challenge, the majority explained, proceeds in two steps. A court must first determine the full scope of the law, identifying every covered application, and then decide which applications violate the First Amendment and measure those against the applications that survive constitutional scrutiny. Only if the unconstitutional applications substantially outweigh the constitutional ones may the statute be struck down in its entirety.

Neither circuit had performed that analysis, the majority held, because both had concentrated on a narrow slice of the laws’ operation. The arguments and reasoning below focused on how the statutes applied to the flagship content feeds of the largest services: Facebook’s News Feed, YouTube’s homepage, and near equivalents. But the statutes were drafted broadly enough to cover functions far beyond the curated feed. They reached email services, direct messaging, and other tools through which platforms transmit user communications, functions in which the constitutional questions may look very different from those presented by a public feed. By analyzing only the feed applications, the lower courts had not assessed the full range of covered activity, and without that assessment no court could properly weigh the constitutional against the unconstitutional applications. The opinion also noted that the laws contained distinct kinds of obligations that required separate treatment: the content-moderation restrictions, which tell platforms what speech they must or must not carry, and the individualized-explanation requirements, which compel platforms to give users detailed reasons for particular moderation decisions. Those disclosure mandates raised their own First Amendment questions, because a compelled statement of reasons is itself a burden on the speaker’s expressive choices, and the lower courts had not measured those applications either. Because the Court sits as a court of review rather than of first view, it could not undertake the needed inquiries itself on the existing records, and the cases had to return to the circuits for the full facial analysis.

That procedural disposition, however, did not mean the majority said nothing about substance. Anticipating the analysis the lower courts would perform on remand, Justice Kagan’s opinion set out the First Amendment framework that applies to the core feed applications. The opinion stated that a platform’s curation of the content it displays, its choices about which third-party messages to include, exclude, filter, alter, or label, is itself expressive activity protected by the First Amendment. In reaching that conclusion, the majority drew on the editorial-discretion precedents that have protected private speakers’ decisions about the speech they present. Miami Herald Publishing Co. v. Tornillo, decided in 1974, held that the government could not compel a newspaper to print a reply to its editorials, protecting the editorial function of selecting and presenting material. Pacific Gas and Electric Co. v. Public Utilities Commission of California, decided in 1986, extended that protection to a utility’s decisions about what to include in its newsletter, rejecting compelled carriage of messages the private party did not wish to endorse. Hurley v. Irish-American Gay, Lesbian and Bisexual Group of Boston, decided in 1995, held that organizers of a parade could not be forced to include a group whose message they did not wish to convey, because the composition of the parade was itself expression. The majority applied that line of authority to the platforms’ feeds: the choices that shape a feed give it a particular expressive quality and constitute the exercise of protected editorial control. On that reasoning, a state effort to force platforms to carry speech they would prefer to remove is a form of government interference with private expression, not a neutral regulation of conduct.

The opinion went further and addressed the interest Texas had asserted for its law. The state had argued that its statute served the goal of balancing the ideological composition of the major feeds. The majority responded that a state’s desire to change the balance of speech on the platforms’ feeds is not the kind of important or substantial interest that could justify restricting the platforms’ expression under any level of scrutiny. A state, the opinion stated, may not interfere with private actors’ speech to advance its own vision of ideological balance. That passage supplied the doctrinal foundation for the proposition at the heart of this section: because the curation of third-party content is itself expressive activity under the First Amendment, withdrawing a liability statute would not by itself license the government to compel platforms to carry speech they wish to exclude. The repeal of a liability shield and the imposition of a must-carry obligation are constitutionally distinct operations, and the latter must be measured against the editorial-discretion precedents regardless of the former. The majority’s reasoning on this point drew strength from a simple structural observation. The editorial precedents all involved private parties who assembled the speech of others into a product that bore their own expressive choices, and the feed performs that assembly function at scale. The fact that the platforms host vastly more speech than a newspaper or a parade did not, in the majority’s analysis, change the character of the curation choice itself: selecting which messages to feature and which to suppress remains an exercise of expressive judgment about what a platform’s space should communicate.

The procedural posture bears emphasis. The Court did not issue final merits rulings on the full statutes. It vacated the judgments of the Eleventh and Fifth Circuits and remanded for the complete facial analysis, leaving the preliminary injunctions against both laws in place in the interim. The majority expressly left open questions that the lower courts had not developed, including the states’ common-carrier arguments, which contend that the platforms are so central to public discourse that they may be subjected to must-carry obligations of the kind historically applied to telephone companies and other conduits. The opinion also left the deeper history of must-carry regulation to the lower courts, neither endorsing nor foreclosing the claim that the platforms’ functions could be analogized to earlier communications intermediaries. Everything about the substantive application of the First Amendment to the laws’ full range of covered functions remained for the remand proceedings.

The Separate Opinions

The separate opinions, while joining the judgment to vacate and remand, divided over how much the majority should have said about the merits. Justice Barrett, who joined the majority in full, wrote a concurrence emphasizing the practical difficulty of the task the facial challenges presented. A statute that covers a broad swath of varied platforms and functions, she observed, makes a facial challenge a daunting undertaking, because the constitutional analysis must account for applications that differ substantially from one another. Her concurrence functioned as a caution to litigants and lower courts: the breadth of the laws’ coverage requires careful attention to each application rather than sweeping assumptions about the covered services. Where the majority treated the feeds as the paradigmatic application around which to organize the remand guidance, Barrett’s opinion underscored that the feeds may be only one corner of a much larger statutory map, and that a court which mistakes the corner for the map cannot sustain a facial judgment either way.

Justice Jackson concurred in part and in the judgment, joining Parts I, II, and III-A of the majority opinion but declining to join the portions that opined on the constitutionality of the laws as applied to the main feeds. Her opinion stressed judicial restraint, arguing that the Court should avoid deciding more than was necessary to resolve the cases before it. Because the disposition turned on the inadequacy of the facial analysis below, the merits discussion of the feeds was, in her view, an unnecessary step, and the Court ought to have left those questions entirely to the remand.

Justice Thomas wrote an opinion concurring in the judgment, questioning whether federal courts possess the authority to entertain facial challenges at all in this posture, and echoing the concern that courts should not rule on the constitutionality of entire statutes on underdeveloped records.

The fourth separate opinion, by Justice Alito and joined by Justices Thomas and Gorsuch, also concurred in the judgment only, and it took direct issue with the majority’s substantive discussion. Justice Alito’s opinion cautioned against analogizing platforms to traditional media without sufficient factual development, arguing that when constitutional requirements are applied to new technology, courts should proceed with caution and build the record first. On the merits of the facial challenges, the opinion concluded that the challengers had not yet demonstrated that the laws were facially unconstitutional, while acknowledging that further proceedings might alter that assessment. It described the majority’s application of First Amendment principles to the feeds as nonbinding dicta, statements unnecessary to the judgment and therefore not precedential. Underlying the disagreement was a different view of what the platforms do when they moderate: the Alito opinion treated the analogy between platform curation and the editorial judgments of newspapers and parade organizers as contestable rather than settled, particularly given how little the records established about how the various covered functions actually operated.

Murthy v. Missouri and the Standing Barrier

Five days before the NetChoice decisions, on June 26, 2024, the Court decided a companion question in internet law from the opposite direction: not whether the government could compel platforms to carry speech, but whether the government had unconstitutionally pressured platforms to remove it. Murthy v. Missouri, originally filed as Missouri v. Biden, arose from communications during the first Trump and Biden administrations between federal officials and major platforms. The plaintiffs, the states of Missouri and Louisiana along with five individuals, alleged that officials including White House staff, the Surgeon General, the Centers for Disease Control and Prevention, the Federal Bureau of Investigation, and the Cybersecurity and Infrastructure Security Agency had coerced platforms to suppress content concerning the 2020 election and the COVID-19 pandemic. Internal communications obtained in discovery showed officials flagging specific posts for removal or demotion, sometimes expressing frustration when platforms did not act and at times hinting at regulatory consequences such as antitrust scrutiny or changes to the liability statute. The plaintiffs contended that this conduct amounted to jawboning, and that it transformed the platforms’ moderation decisions into state action subject to First Amendment limits.

In internet law doctrine, jawboning refers to informal government pressure on private intermediaries, a practice distinct from direct regulation. The government does not enact a rule or issue an order; instead, officials use communications, threats, or implied consequences to induce a private actor to restrict speech the government could not restrict itself. The governing test for such informal pressure traces to Bantam Books, Inc. v. Sullivan, decided in 1963. There, a Rhode Island commission charged with encouraging morality in youth sent notices to book distributors identifying publications as objectionable, circulated lists of such publications to police departments, and requested the distributors’ cooperation in withholding them from sale. The commission possessed no formal enforcement power, but the Court held that its thinly veiled threats of prosecution, notices phrased virtually as orders, and follow-up police visits constituted a scheme of state censorship effectuated through extralegal sanctions, violating the Fourteenth Amendment. The doctrine drawn from that decision holds that when government suggestions are tantamount to commands, and a private party’s compliance is not genuinely voluntary, the resulting suppression is attributable to the state. The line between permissible persuasion and unconstitutional coercion turns on whether the government’s communications carried a coercive threat that the intermediary could not reasonably disregard.

The Court in Murthy did not apply that test. In a 6-3 decision, Justice Barrett writing for the majority and joined by Chief Justice Roberts and Justices Sotomayor, Kagan, Kavanaugh, and Jackson, the Court held that none of the plaintiffs had established Article III standing to seek an injunction against any defendant, and therefore the Court lacked jurisdiction to reach the coercion merits. The decision arrived only weeks after the Court had clarified the coercion framework itself in National Rifle Association of America v. Vullo, decided in May 2024, where the Court reaffirmed that government officials may not use the power of their office to punish or suppress disfavored expression through pressure on third parties, and that such coercion violates the First Amendment even when it operates through informal channels rather than formal enforcement. Vullo supplied the substantive doctrine; Murthy showed how rarely that doctrine reaches the merits when the plaintiffs cannot connect a particular removal to a particular official act. The standing analysis turned on traceability and redressability. The majority found that the record lacked specific causation findings linking any discrete instance of content moderation to a particular government communication. While the evidence showed that officials had played a role in at least some moderation choices, the platforms had independent incentives to moderate content, maintained their own pre-existing policies against misinformation, and often exercised their own judgment in deciding what to remove or demote. That independence is central to internet law’s treatment of jawboning claims, because a platform that would have removed a post under its own rules presents a causation problem the plaintiff cannot solve by pointing to a government email about the same post. The platforms’ moderation operations predated the officials’ outreach, scaled with the companies’ own commercial and reputational interests, and continued after the outreach changed in character, which meant the record could not isolate the government’s contribution from the companies’ own. The Fifth Circuit, the majority said, had glossed over these complexities by attributing every platform decision at least in part to the defendants. Without a showing that a specific government demand caused a specific removal, the past injury was not fairly traceable to the officials’ conduct. The plaintiffs also failed the forward-looking requirement for injunctive relief, because they could not show a substantial likelihood that they would again suffer injury traceable to the government in the future, particularly given changes in the platforms’ own policies and in the personnel and practices of the agencies involved. That forward-looking requirement is what distinguishes a damages suit over a past removal from a request that a court order officials to stop communicating with platforms altogether, and it is why standing doctrine so often decides jawboning cases before coercion doctrine gets a hearing. Even if the past contacts looked coercive, an injunction issues only where the plaintiff shows that the coercion is likely to recur and to harm that plaintiff, a showing the majority found entirely speculative on the record presented.

Justice Alito dissented, joined by Justices Thomas and Gorsuch, and his opinion viewed both the standing and the merits questions differently. On standing, the dissent argued that the record established the necessary link for at least one individual plaintiff, whose content had been suppressed in direct response to government communications, and that the threat of future pressure was sufficiently concrete to support injunctive relief. On the merits, the dissent described the evidence as showing a coordinated campaign of pressure by the federal government to suppress disfavored viewpoints, conduct that, in the dissent’s view, clearly crossed the line from persuasion to coercion under the Bantam Books framework. The dissent treated the platforms’ actions as the product of sustained government demands rather than independent editorial judgment, and it would have held the conduct unconstitutional.

The result left the jawboning doctrine in an unresolved state. Because the majority resolved the case on standing, it expressly took no view on whether the officials’ communications violated the First Amendment, and the lower courts’ injunctions against the government were reversed. The Bantam Books test remains the governing framework for evaluating informal government pressure on intermediaries, but Murthy established that such claims must first survive a demanding standing inquiry: plaintiffs must tie a specific moderation action to a specific government communication, rather than relying on a general pattern of contact, and must show that future harm is likely absent judicial intervention. For internet law, the decision clarified that the constitutional question in jawboning cases is almost always preceded by a causation question, and that the platforms’ independent moderation incentives and policies make that causation question genuinely difficult to answer on a thin record.

The two strands of the 2024 term, taken together, map the moderation line from both sides. The NetChoice decisions hold that direct state compulsion of carriage must answer to the First Amendment’s protection of editorial curation, while remanding the question of how that principle applies across the many functions the state laws reach. Murthy holds that indirect federal pressure on moderation must answer to the coercion doctrine of Bantam Books, while requiring plaintiffs to prove that the government’s hand, rather than the platform’s own judgment, caused the removal. In both lines, the Court of 2024 avoided final substantive rulings where procedure supplied the narrower ground, and in both lines the questions that matter most, how far a state’s common-carrier argument can go and where persuasion ends and coercion begins in the government’s dealings with intermediaries, were preserved for future cases with fuller records.

The Absent Line: The Liability Provision the Court Never Construed

Few statutes decide more cases while receiving less authoritative interpretation. Section 230 of the Communications Decency Act of 1996, codified at 47 U.S.C. 230, supplies the defense that ends most lawsuits against online intermediaries at the pleading stage. Its central clause provides that no provider or user of an interactive computer service shall be treated as the publisher or speaker of information provided by another information content provider. Federal and state courts have applied that language to dismiss claims running from defamation to product liability, building an immense body of doctrine almost entirely without guidance from the Supreme Court. The provision is the most litigated rule in internet law, and no opinion of the Court construes it.

The omission was nearly repaired. In the October 2022 Term the Court granted review in a case that squarely presented the scope of the immunity, Gonzalez v. Google LLC, and paired it with a companion dispute, Twitter, Inc. v. Taamneh, that raised only the merits of the underlying claims. On May 18, 2023, the Court decided the companion case and used its result to dispose of the immunity case without touching the immunity question. The maneuver was procedurally tidy and doctrinally consequential: the liability provision survived another Term unexamined, and the lower court consensus that Justice Clarence Thomas had spent three years inviting the Court to revisit remained the law by default rather than by decision.

The edifice that consensus built is worth picturing before examining the moments the Court almost dismantled it. The foundation is Zeran v. American Online, Inc., 129 F.3d 327 (4th Cir. 1997), which read subsection (c)(1) to bar any claim that would hold a service liable for exercising a publisher’s traditional editorial functions, including decisions whether to publish, withdraw, postpone, or alter third party content, and which rejected distributor liability even with notice of unlawful material. Every regional circuit to consider the question adopted Zeran’s framework. The leading exception that proved the rule came from the Ninth Circuit sitting en banc in Fair Housing Council of San Fernando Valley v. Roommates.com, LLC, 521 F.3d 1157 (2008), which held that a housing website that required users to answer discriminatory questions about sex, sexual orientation, and family status was itself an information content provider for those answers. Roommates.com drew the content development line more aggressively than any decision before or since, and later panels cabined it nearly to its facts. Together these opinions, plus thousands of unpublished dismissals, constitute the operative law of platform liability in the United States, all of it written without a single controlling Supreme Court construction of the words Congress enacted.

A case presented and a question answered elsewhere

The Paris attacks of November 13, 2015, killed 130 people in coordinated ISIS assaults on restaurants, a stadium, and the Bataclan concert hall. Among the dead was Nohemi Gonzalez, a 23 year old American university student spending a semester in France. Her family sued Google, the owner of YouTube, under the civil liability provisions of the Anti-Terrorism Act, 18 U.S.C. 2333, as amended by the Justice Against Sponsors of Terrorism Act of 2016, known as JASTA. The complaint alleged that YouTube recommendation algorithms steered viewers toward ISIS recruitment videos, amplifying the organization’s reach, and that Google shared advertising revenue with ISIS through its AdSense program. The district court dismissed the suit. The Ninth Circuit, in Gonzalez v. Google LLC, 2 F.4th 871 (2021), held that the recommendation claims were barred by Section 230, reasoning that recommending third party videos was a traditional publisher function, while leaving a narrow path open on the revenue sharing allegations. The family petitioned for certiorari, and the Court agreed to decide whether Section 230 barred claims alleging that recommendation algorithms facilitated terrorism. For the first time, the scope of the immunity itself was before the Court.

The companion case arose from the January 1, 2017, ISIS attack on the Reina nightclub in Istanbul, in which a gunman killed 39 people, including Nawras Alassaf, a Jordanian citizen. His family sued Twitter, Facebook, and Google under JASTA’s civil aiding and abetting provision, 18 U.S.C. 2333(d)(2). The plaintiffs alleged that for years the companies had knowingly allowed ISIS and its supporters to use their services and recommendation algorithms as tools for recruiting, fundraising, and spreading propaganda, and had profited from advertisements placed alongside ISIS content. The district court dismissed the complaint for failure to state a claim. The Ninth Circuit reversed and allowed the case to proceed. The companies sought review on the JASTA merits alone, presenting no Section 230 question, and the Court granted it. The two cases were argued on consecutive days in February 2023, with Taamneh supplying the vehicle for the underlying liability question and Gonzalez supplying the vehicle for the immunity question. At the Gonzalez argument, several Justices voiced open unease about the judiciary rewriting the provision. Justice Elena Kagan observed that the Court was not composed of the nine greatest experts on the internet, a remark that captured the bench’s evident reluctance to redesign, from first principles, a liability regime on which the online economy had been built. Counsel for the plaintiffs urged that recommendation algorithms were not publishing at all but the platforms’ own affirmative conduct, while counsel for Google warned that stripping immunity from recommendations would expose every search engine, news feed, and thumbnail gallery to the same theory. The argument ended without any clear majority for either side’s reading of the text, which made the eventual disposition, avoidance through Taamneh, all the more telling: the Court chose not to decide what it could not confidently construe.

Twitter, Inc. v. Taamneh, 598 U.S. 471 (2023), was decided unanimously, with Justice Thomas writing for the Court. JASTA imposes secondary civil liability on anyone who aids and abets, by knowingly providing substantial assistance, an act of international terrorism. Congress had identified Halberstam v. Welch, 705 F.2d 472 (D.C. Cir. 1983), as supplying the proper legal framework for that liability, and the opinion built its analysis on that foundation. Aiding and abetting, the Court held, requires conscious, voluntary, and culpable participation in another’s wrongdoing, and the defendant must have aided the terrorist act that injured the plaintiffs rather than terrorism in the abstract. The companies’ alleged conduct did not meet that standard. Creating global communications services and designing algorithms that match content with users likely to view it, the opinion explained, was not culpable participation in the Reina attack. The algorithms were agnostic as to the nature of the content, matching ISIS videos to interested viewers in the same way they matched anything else to anyone else. The companies’ relationship with ISIS was arm’s length, passive, and indifferent, identical to their relationship with more than a billion other users. The failure to remove ISIS accounts and content was nonfeasance, not assistance, and the opinion compared the services to cell phones, email, and the internet generally: communications infrastructure that wrongdoers happen to use. Justice Ketanji Brown Jackson concurred separately to emphasize that other cases, presenting different allegations and different records, could lead to different conclusions.

Gonzalez was decided the same day in a brief per curiam opinion, Gonzalez v. Google LLC, 598 U.S. 617 (2023). Because the Gonzalez complaint rested on substantially the same theory of platform responsibility that Taamneh had just rejected, the Court saw no need to reach the immunity defense at all. The opinion observed that the complaint, independent of Section 230, appeared to state little if any plausible claim for relief, noted the Ninth Circuit’s unchallenged holdings below, and declined to address the statute’s application. The Court vacated the Ninth Circuit’s judgment and remanded for reconsideration in light of Taamneh. The Section 230 question on which certiorari had been granted went unanswered by the very Court that had agreed to answer it.

What the Court did not decide is easier to list than what it did. It said nothing about whether algorithmic recommendations fall within the publisher or speaker protection of 47 U.S.C. 230(c)(1). It said nothing about the relationship between subsection (c)(1) and the conditional good faith immunity of subsection (c)(2). It said nothing about when a service that solicits, ranks, and promotes user submissions becomes an information content provider responsible for content development. The Ninth Circuit’s holding that recommending videos is a protected publisher function survived, but only as circuit precedent, carrying none of the Court’s imprimatur. This is the liability provision the Court has not construed: the rule that disposes of more internet law litigation than any other, still awaiting its first authoritative reading. The episode carried a quiet irony. Thomas, who wrote the unanimous Taamneh opinion that sidestepped Section 230, was also the Justice who had spent the preceding three years urging the Court to confront it.

Denials that signaled without deciding

The invitation had been issued twice before, in cases the Court declined to hear. Force v. Facebook, Inc., 934 F.3d 53 (2d Cir. 2019), was brought by the families of five Americans killed or injured in Hamas attacks in Israel. The plaintiffs alleged that Facebook had provided material support to Hamas by hosting its content and by steering users toward it through friend suggestions and content recommendations. The Second Circuit held that Section 230 barred the claims, becoming the first federal appellate court to apply the statute to civil terrorism claims, and reasoned that recommender algorithms were essentially neutral tools performing a distributor function rather than publishing. Chief Judge Robert Katzmann wrote a lengthy opinion concurring in part and dissenting in part, arguing that mounting evidence of platforms designing algorithms to maximize engagement by pushing users toward congenial extremist content could take recommendation outside traditional publishing altogether. His dissent warned of systems nudging susceptible users ever further down dark paths. The Supreme Court denied certiorari in 2020, Force v. Facebook, 140 S. Ct. 2761, without any separate writing.

The first separate statement came later that year. Malwarebytes, Inc. v. Enigma Software Group USA, LLC, 141 S. Ct. 13 (2020), began as a dispute between competing cybersecurity vendors. Enigma alleged that Malwarebytes had configured its software to flag Enigma’s programs as threats in order to suppress competition. Malwarebytes invoked Section 230(c)(2), which shields good faith acts to block or filter objectionable material, and the Ninth Circuit denied immunity on policy and purpose grounds tied to the alleged anticompetitive conduct. The Court denied certiorari on October 13, 2020. Thomas again agreed with the denial and again wrote separately, this time at length. Courts, he wrote, had long emphasized nontextual arguments when interpreting the statute, leaving questionable precedent in their wake. They had discarded the common law distinction between publisher liability and distributor liability. They had extended immunity to companies for their own content despite subsection (c)(1)’s limitation to information provided by another information content provider. And by reading subsection (c)(1) capaciously, they had eviscerated the narrower, conditional shield that Congress had written into subsection (c)(2). He closed by saying that in an appropriate case, it behooved the Court to take up the question. Commentators read the statement as an open invitation, and litigants treated Gonzalez as its answer. When the Court granted review in Gonzalez, the invitation appeared accepted. The per curiam disposition returned it unopened.

Thomas returned to the subject in 2022. In Doe v. Facebook, Inc., 595 U.S. ___ (2022), the Court again denied certiorari in a case presenting the scope of the platform immunity, and Thomas again filed a statement respecting the denial, reiterating that the Court should take up the statute’s meaning in an appropriate case. With that, the invitation had been issued twice.

These statements deserve precise characterization, because their procedural posture determines their legal weight, which is none. A statement respecting the denial of certiorari is not a holding. It is not even dissent, since there is no decision to dissent from; the denial itself decides nothing on the merits and binds no court. Thomas’s statements were joined by no colleague, cited by no majority, and adopted by no lower court as authority. Their significance was entirely predictive and hortatory: one Justice signaling that the lower court consensus on Section 230 rested on foundations he considered questionable, and inviting future litigants to bring a suitable case. Gonzalez looked like the suitable case. The Court’s refusal to decide it there confirmed that an invitation, however pointed, is not a grant.

The open questions

A future construction of the provision would have to resolve at least four textual puzzles, each of which remains open precisely because the Court has never addressed it.

The first concerns the relationship between the statute’s two operative subsections. Subsection (c)(1) is phrased as a rule of treatment rather than a grant of immunity: no provider or user of an interactive computer service shall be treated as the publisher or speaker of information provided by another information content provider. Subsection (c)(2) grants immunity in terms, but only for defined conduct: good faith acts to restrict access to objectionable material, and the provision of technical means to filter such material. The open question is whether subsection (c)(1) does independent immunizing work beyond subsection (c)(2), or whether the lower courts, beginning with Zeran v. American Online, Inc., 129 F.3d 327 (4th Cir. 1997), stretched a definitional clause into a general immunity so broad that subsection (c)(2)’s good faith limitation became surplusage. Canons of construction disfavor readings that render enacted language meaningless, yet the prevailing doctrine gives subsection (c)(1) nearly all the work and subsection (c)(2) almost none. Only the Court can say whether that allocation follows the text. The history behind the two subsections sharpens the puzzle. Congress wrote Section 230 to overrule Stratton Oakmont, Inc. v. Prodigy Services Co., a 1995 New York trial court decision that had treated an online service as a publisher, and therefore liable for user defamation, precisely because the service screened and moderated content. The statute’s caption, Protection for Good Samaritan blocking and screening of offensive material, reflects that purpose: subsection (c)(2) was the operative shield for services that removed objectionable material in good faith, while subsection (c)(1) ensured that hosting third party content did not by itself make a service the content’s publisher. On that original design, the two provisions divide the labor, with (c)(1) addressing status and (c)(2) addressing conduct. The Zeran line collapsed the distinction by giving (c)(1) the force of a general immunity, leaving (c)(2) with little left to protect. Whether that collapse was faithful to the design is the kind of question only an authoritative construction can settle.

The second concerns algorithmic recommendation, the very issue Gonzalez presented. Zeran defined publisher functions as the traditional editorial decisions whether to publish, withdraw, postpone, or alter content. Recommendation engines do not publish in that sense; they select, rank, and amplify material that others have already published, often without human review of any particular item. The open question is whether that activity counts as publishing and therefore falls within the statutory bar, or whether it constitutes the platform’s own conduct in developing and presenting content, which the statute does not protect. The Force majority treated recommenders as neutral distributor tools, while Katzmann’s partial dissent treated engagement optimized amplification as the company’s own act. Both readings claim support in the same statutory words. Taamneh’s observation that the algorithms were agnostic as to content addressed culpability under the Anti-Terrorism Act, not immunity under Section 230, and it binds no court on the latter question.

The third concerns distributor liability and knowledge. At common law, distributors such as bookstores and libraries faced liability for unlawful content only with knowledge of its character, a middle tier between the strict responsibility of publishers and the immunity of mere conduits. Thomas’s Malwarebytes statement suggested that Congress legislated against that background and may have preserved the publisher distributor distinction rather than abolishing it. The open question is whether a service with specific knowledge that it hosts unlawful material, which nevertheless leaves the material in place, remains a distributor outside subsection (c)(1)’s bar. Every circuit to consider the question has followed Zeran in rejecting knowledge based distributor liability. The Supreme Court has never said whether Zeran was right, and its silence leaves a foundational premise of internet law resting on a single 1997 circuit opinion.

The fourth concerns the statute’s definitional boundaries. The immunity protects providers and users of interactive computer services, a capacious category, but withholds protection from information content providers, defined as persons responsible in whole or in part for the creation or development of information. The open question is where one category ends and the other begins when a service solicits submissions, structures their presentation, prompts users for particular kinds of content, and promotes selected items to wider audiences. Lower courts have generally treated even substantial editorial shaping as falling short of content development, but the line has never been drawn by the Court, and litigants continue to test it in cases involving structured data, algorithmic promotion, and platform designed features. Roommates.com remains the high water mark of the aggressive reading: Chief Judge Alex Kozinski’s en banc opinion held that by requiring users to select from prepopulated discriminatory answers, the website became a developer of the resulting profiles, not a passive transmitter of them. Later decisions distinguished that holding on the ground that drop down menus and search tools merely organize user supplied content rather than create it, but the distinction is asserted more often than it is explained. Each new product feature, from algorithmic feeds to generative summaries, reopens the same inquiry under a precedential regime that has never received the Court’s endorsement.

The significance of these unanswered questions becomes clear when set against what the Court has decided. Taamneh and Gonzalez resolved the standard for civil aiding and abetting liability under JASTA: conscious, voluntary, and culpable participation in the specific terrorist act, not merely the operation of a communications service that wrongdoers also use. Those are the claims these decisions resolve, and their holdings matter for terrorism litigation. They do not interpret the liability provision that governs the far larger run of cases against platforms. Because that provision stands unconstrued, the binding rules determining what governments may require of platforms have come from elsewhere, principally First Amendment doctrine. In Moody v. NetChoice, LLC, 603 U.S. 707 (2024), the Court addressed state laws regulating platform content moderation through the lens of editorial discretion and compelled speech. In Murthy v. Missouri, 603 U.S. 43 (2024), it addressed government pressure on platform content decisions through the lens of standing and state action. Those decisions set constitutional boundaries on the state’s relationship to platforms. They do not construe the statute that decides most private suits against platforms at the threshold.

The practical consequence is a litigation landscape shaped by avoidance. Plaintiffs seeking recovery for online harms have learned to plead around the unconstrued provision, framing claims as product liability, negligent design, or failure to warn rather than as attacks on publishing decisions, hoping to land on the Roommates.com side of a line no higher authority has fixed. Defendants invoke the Zeran consensus at the pleading stage and usually prevail, but each close case turns on which circuit’s gloss applies and how aggressively a given panel reads the content development exception. The result is precisely the regime Thomas described: broad, judge made, and resting on policy and purpose as much as on text. A Supreme Court construction would not necessarily narrow the immunity; it could just as easily confirm the lower courts’ reading and entrench it. What it would supply, either way, is an answer attributable to the statute’s words rather than to three decades of accretion.

The asymmetry defines the field. Constitutional law governs what the government may demand of intermediaries, while the statute governing what private plaintiffs may demand of them awaits its first authoritative reading. Thousands of dismissals each year rest on circuit precedent, on Zeran and its progeny, on dicta and invitations that never ripened into holdings. The most litigated provision in internet law remains, after the closest approach in its history, the line the Court has never drawn.

The Doctrine Ledger

The holdings discussed in this article are collected here for reference. Each row names the decision, the statute or state law at issue, the holding in one sentence, and what the decision established for online speech.

Decision	Year	Statute or state law at issue	Holding in one sentence	What it established for online speech
Reno v. ACLU	1997	Communications Decency Act, indecency provisions, 47 U.S.C. 223	Struck down the online indecency provisions and held that internet speech receives the full First Amendment protection of print rather than the diminished protection of broadcast.	Fixed the constitutional default of the network: strict scrutiny for content-based regulation and the print analogy as the medium’s doctrinal home.
Ashcroft v. ACLU I/II and COPA	2002/2004	Child Online Protection Act of 1998, criminal penalties for commercial communications harmful to minors	Sustained the preliminary injunction against COPA because filtering software offered a less restrictive alternative to criminal punishment of publishers.	Confirmed that direct criminal prohibitions on protected online expression face strict scrutiny and usually fail where less restrictive alternatives exist.
United States v. American Library Association	2003	Children’s Internet Protection Act of 2000, condition on E-rate discounts and library grants	Upheld CIPA against a facial challenge as a permissible condition on a federal subsidy rather than a penalty on speech.	Drew the funding-condition line: Congress may attach strings to money it need not spend, while direct prohibitions face the strictest judicial scrutiny.
Packingham v. North Carolina	2017	N.C. Gen. Stat. section 14-202.5, felony for registered sex offenders to access commercial social networking sites	Struck down the blanket ban on social media access because it was not narrowly tailored to the state’s interest in protecting minors.	Made access to social networking platforms constitutionally significant, permitting regulation of conduct online but not blanket exclusion from the forum.
Twitter v. Taamneh and Gonzalez v. Google	2023	JASTA civil liability, 18 U.S.C. 2333; Section 230 presented in Gonzalez but not decided	Set the aiding-and-abetting standard for terrorism claims against platforms and disposed of Gonzalez without construing Section 230.	Defined the terrorism-liability standard while leaving the most litigated liability provision in internet law without a Supreme Court construction.
Moody v. NetChoice / NetChoice v. Paxton	2024	Texas H.B. 20 and Florida S.B. 7072, state regulation of platform content moderation	Vacated and remanded for a complete facial-challenge analysis while stating that a platform’s feed curation is expressive activity protected by the First Amendment.	Held that compelled carriage on curated feeds must answer to the editorial-discretion precedents, and that a state’s desire for ideological balance is not a sufficient justification.
Murthy v. Missouri	2024	No statute; informal federal pressure on platform moderation decisions, known as jawboning	Held that the plaintiffs lacked Article III standing to seek an injunction, taking no view on whether the government’s conduct violated the First Amendment.	Established that jawboning claims must first survive a demanding causation and standing inquiry before the Bantam Books coercion test can be applied.

The Constitution, Not the Statute

The namable claim of this article is a single sentence: the rules that determine what governments may require of platforms come from First Amendment doctrine, because the liability statute has never been construed by the Court. Every major dispute between the state and the intermediary in this account was resolved through constitutional analysis. Reno applied strict scrutiny to content-based regulation and chose the print analogy. Ashcroft v. ACLU II applied strict scrutiny again and sustained the injunction because filtering software supplied a less restrictive alternative. United States v. American Library Association turned on the spending power and the unconstitutional-conditions doctrine, not on any interpretation of the liability provision. Packingham tested the tailoring of a speaker-based exclusion from a forum. Moody v. NetChoice reached for the editorial-discretion precedents to measure compelled carriage. Murthy v. Missouri was decided on Article III standing before the coercion test could be applied. Section 230, the statute that disposes of more private litigation than any other provision in the field, contributed nothing to these holdings. A lawyer forecasting a clash between government and platform therefore studies constitutional doctrine, while a lawyer forecasting a clash between private parties studies the lower court consensus that has grown up without the Court.

The Series Thesis: Legislation Governs Only Where Courts Let It

This article belongs to a series that examines how legislation works in the United States: how statutes are written, how they are tested, and what survives contact with the courts. The internet-law installment tells a story in which legislation repeatedly failed to do the work Congress assigned it. The indecency provisions of the Communications Decency Act fell to the First Amendment. Its successor, COPA, never escaped preliminary relief and died without being enforced against a single speaker. CIPA survived only by shrinking into a condition on a subsidy, a modest instrument for the ambition Congress had begun with. The one provision that endured, section 230, endures through construction by lower courts rather than through any authoritative reading by the Court. The parallel with structural media regulation is instructive: where Congress once built comprehensive carriage frameworks for earlier technologies, the internet statutes aimed at speech itself could not survive constitutional review. The series thesis, applied here, is that legislation does not govern a field simply by being enacted; it governs only where the Constitution permits and where courts are willing to construe it.

Closing

The field, then, remains open at both ends. The constitutional lines are drawn: full protection for online expression, funding conditions distinguished from prohibitions, access treated as constitutionally significant, compelled carriage measured against editorial discretion, and informal pressure tested for coercion only after standing is satisfied. The statutory line is still missing, and its absence is the largest fact about American internet law. Until the Court construes the provision that decides most private cases, the field will continue to run on constitutional holdings and circuit accretion, a regime that functions after its fashion but answers to no single text. The contrast with earlier media regulation, where Congress built a durable structural framework as traced in the earlier media structural regulation framework, only sharpens the point: here the statute that was supposed to govern awaits its first reading. Readers who wish to trace the enacted provisions discussed here against their legislative history may consult the VaultBook legislation study notebook.

Frequently Asked Questions

Q: Does the First Amendment apply fully to the internet under US internet law?

Yes. In Reno v. American Civil Liberties Union, 521 U.S. 844 (1997), the Supreme Court held that expression on the internet receives the full First Amendment protection given to print, not the diminished protection the Court had tolerated for broadcast. The Government had urged the Court to apply the reduced-scrutiny framework of FCC v. Pacifica Foundation, but Justice Stevens rejected the analogy because the internet lacked broadcasting’s history of pervasive regulation, its spectrum scarcity, and its invasive character. Since internet users must take affirmative steps to retrieve content, the medium resembled a library shelf more than a radio dial. That holding made content-based regulation of online speech subject to strict scrutiny, the most demanding constitutional test, and it has remained the default premise of internet law ever since.

Q: What did Reno v. ACLU establish in internet law?

Reno struck down the indecency provisions of the Communications Decency Act and fixed the constitutional status of the network itself. The decision supplied three linked propositions: first, that internet expression gets the full First Amendment protection of print; second, that a statute’s constitutionality turns on factual findings about how the medium actually operates, with the district court’s description of a decentralized, non-invasive network doing real doctrinal work; and third, that the line between unprotected obscenity and protected indecent expression would be held firmly, with criminal penalties on indecent but non-obscene material failing both vagueness and overbreadth analysis. Every later controversy over online speech has been argued within the framework Reno established.

Q: What did Packingham decide about internet law?

Packingham v. North Carolina, 582 U.S. 98 (2017), struck down a state statute that made it a felony for registered sex offenders to access commercial social networking websites. Justice Kennedy’s majority opinion held that the law was not narrowly tailored to the state’s interest in protecting minors even under intermediate scrutiny, because it punished logging in rather than predatory conduct. The decision treated access to social media as constitutionally significant, describing cyberspace as among the most important places for the exchange of views. Justice Alito concurred in the judgment on narrower grounds, warning against reading the majority’s broad language as a general theory of the internet. The holding redirected regulation from the doorway of the forum to the conduct within it.

Q: Has the Supreme Court ever ruled on section 230 in internet law?

No. Section 230 of the Communications Decency Act, codified at 47 U.S.C. 230, is the most litigated provision in internet law and has never received an authoritative construction from the Supreme Court. The closest approach came in Gonzalez v. Google LLC, 598 U.S. 617 (2023), in which the Court granted review to decide whether the provision barred terrorism claims based on recommendation algorithms, then disposed of the case through its companion decision without reaching the immunity question. Twice before, in Malwarebytes, Inc. v. Enigma Software Group USA, LLC, 141 S. Ct. 13 (2020), and in Doe v. Facebook, Inc., 595 U.S. ___ (2022), Justice Thomas filed statements respecting the denial of certiorari urging the Court to take up the provision in a suitable case. The operative law of platform liability therefore rests on circuit precedent, beginning with Zeran v. American Online, Inc., 129 F.3d 327 (4th Cir. 1997).

Q: Is content moderation protected speech under internet law?

Under the framework the Court set out in Moody v. NetChoice, LLC, 603 U.S. 707 (2024), the answer is yes for the core applications. Justice Kagan’s majority opinion stated that a platform’s curation of the content it displays, its choices about which third-party messages to include, exclude, filter, alter, or label, is expressive activity protected by the First Amendment. The opinion drew on the editorial-discretion precedents of Miami Herald v. Tornillo, Pacific Gas and Electric v. Public Utilities Commission, and Hurley v. Irish-American Gay, Lesbian and Bisexual Group of Boston, which protect private parties’ decisions about the speech they present. The Court expressly left open how that principle applies across the many other functions the state laws reached, such as email and direct messaging, so the protection is firmest where the curation most resembles traditional editorial judgment.

Q: Can states regulate social media under internet law?

States may regulate some aspects of social media, but the First Amendment constrains both the means and the ends. Packingham v. North Carolina struck down a blanket ban on platform access by an entire class of speakers, while Moody v. NetChoice, LLC, 603 U.S. 707 (2024), held that a state’s effort to compel platforms to carry speech must answer to the protection of editorial discretion, and that a state’s desire to balance the ideological composition of feeds is not a sufficient justification under any level of scrutiny. The Court has, however, left open the states’ common-carrier arguments and narrower measures aimed at specific conduct, such as contacting minors or impersonating children online. The through line is tailoring: regulation tied to genuine harms and precisely drawn may survive, while broad mandates that override editorial judgment may not.

Q: What is jawboning in internet law?

Jawboning refers to informal government pressure on private intermediaries, a practice in which officials use communications, threats, or implied consequences to induce a private actor to restrict speech the government could not restrict itself, rather than enacting a rule or issuing an order. The governing test traces to Bantam Books, Inc. v. Sullivan (1963), in which the Court held that a state commission’s thinly veiled threats against book distributors constituted censorship effectuated through extralegal sanctions. When government suggestions are tantamount to commands and the intermediary’s compliance is not genuinely voluntary, the resulting suppression is attributable to the state. In Murthy v. Missouri, 603 U.S. 43 (2024), the Court did not apply that test because the plaintiffs failed to establish standing, but the Bantam Books framework remains the governing standard for evaluating such claims.

Q: Can libraries be required to filter under internet law?

Yes, as a condition of receiving federal funds. In United States v. American Library Association, 539 U.S. 194 (2003), the Supreme Court upheld the Children’s Internet Protection Act against a facial challenge. The plurality treated the statute as an exercise of the spending power: Congress was not required to subsidize internet access in public libraries at all, and it could define the limits of the program it chose to fund. Because the law imposed no criminal sanction and left libraries free to decline the money and operate unfiltered terminals, it was judged under spending-power doctrine rather than the strict scrutiny reserved for direct prohibitions. Justice Kennedy’s concurrence made the judgment depend on libraries’ ability to disable filtering for adults without significant delay, and the plurality left open as-applied challenges where that was not true in practice.

Q: Why did the Court treat the internet like print rather than broadcast?

The print analogy came from the medium’s factual characteristics as found by the district court in Reno. Broadcast had received reduced First Amendment protection for three reasons: a long history of government licensing, the scarcity of available frequencies, and the uniquely pervasive and invasive character of radio and television, which entered homes uninvited. None of those fit the internet. The network had never been subject to pervasive licensing, its publishing capacity was not scarce in the broadcast sense, and its communications were not invasive because users had to take deliberate steps, typing an address or following a link, to retrieve content. Justice Stevens concluded that the special justifications for diminished broadcast protection did not apply, so the internet belonged in the lineage of the pamphleteer and the newspaper, with strict scrutiny for content-based regulation.

Q: What is the difference between a funding condition and a prohibition?

The difference is between Congress deciding how to spend its own money and Congress deciding what speakers may say under threat of punishment, and it determines the constitutional test. A funding condition, like the Children’s Internet Protection Act’s filtering requirement for libraries receiving federal subsidies, is judged under the spending power and the unconstitutional-conditions doctrine: Congress may define the limits of a program it funds, so long as it imposes no penalty beyond the loss of the subsidy and leaves recipients free to walk away. A prohibition, like the criminal penalties of the Child Online Protection Act, is a direct restriction on protected expression subject to strict scrutiny. The former survived in United States v. American Library Association; the latter never survived its preliminary injunction. Form, not only purpose, controls the analysis.

Q: What happened to COPA after the Supreme Court decisions?

COPA never took effect and died through attrition. After the Supreme Court sustained the preliminary injunction in Ashcroft v. American Civil Liberties Union, 542 U.S. 656 (2004), and remanded for trial, the Eastern District of Pennsylvania held a full bench trial and, in 2007, declared the statute facially unconstitutional and converted the preliminary restraint into a permanent injunction. The Third Circuit affirmed in 2008, agreeing that the law was impermissibly overbroad and that filtering software offered a less restrictive path to the statute’s goals. On January 21, 2009, the Supreme Court denied certiorari, leaving the permanent injunction undisturbed. COPA had been enacted more than a decade earlier and had never been enforced against a single speaker, a fate that contrasts sharply with CIPA, the funding-condition statute enacted two years later, which the Court sustained.

Q: Why did the Court vacate and remand the NetChoice cases instead of deciding them?

The disposition was procedural. NetChoice had brought facial challenges against the Texas and Florida statutes, asserting that each law violated the First Amendment on its face and across all applications. A proper facial challenge requires a court first to determine the law’s full scope, identifying every covered application, and then to weigh the unconstitutional applications against the constitutional ones. The majority held that neither circuit had performed that analysis: both had concentrated on how the statutes applied to the flagship content feeds of the largest services, while the laws were drafted broadly enough to cover email, direct messaging, and other functions where the constitutional questions look different. Because the Supreme Court sits as a court of review rather than of first view, it could not undertake the needed inquiries on the existing records, so it vacated the judgments and remanded, leaving the preliminary injunctions in place.

Q: What are the editorial-discretion precedents?

They are the line of Supreme Court decisions holding that private parties’ choices about the speech they present are themselves protected expression. Miami Herald Publishing Co. v. Tornillo (1974) held that the government could not compel a newspaper to print a reply to its editorials. Pacific Gas and Electric Co. v. Public Utilities Commission of California (1986) extended that protection to a utility’s newsletter, rejecting compelled carriage of messages the private party did not wish to endorse. Hurley v. Irish-American Gay, Lesbian and Bisexual Group of Boston (1995) held that parade organizers could not be forced to include a group whose message they did not wish to convey, because the parade’s composition was itself expression. In Moody v. NetChoice, LLC, 603 U.S. 707 (2024), the majority applied this line to platform feeds, reasoning that selecting which messages to feature and which to suppress is an exercise of expressive judgment about what a platform’s space should communicate.

Q: Would repealing section 230 let the government compel platforms to carry speech?

No, because the repeal of a liability shield and the imposition of a must-carry obligation are constitutionally distinct operations. As the Moody v. NetChoice majority reasoned, a platform’s curation of its feed is expressive activity protected by the First Amendment, drawing on the editorial-discretion precedents that bar the government from compelling private parties to present speech they would prefer to exclude. Removing section 230 would expose platforms to greater private liability for third-party content, but it would not erase the First Amendment barrier to a state law forcing them to carry it. A must-carry mandate would still have to survive the scrutiny applicable to compelled speech, and the Court stated that a state’s desire to balance the ideological composition of feeds is not a sufficient interest under any level of scrutiny.

Q: What is the Bantam Books coercion standard?

Bantam Books, Inc. v. Sullivan (1963) supplies the test for when informal government pressure on an intermediary becomes unconstitutional state action. A Rhode Island commission charged with encouraging morality in youth sent notices to book distributors identifying objectionable publications, circulated lists to police departments, and requested cooperation in withholding them from sale. The commission possessed no formal enforcement power, but the Court held that its thinly veiled threats of prosecution, notices phrased virtually as orders, and follow-up police visits constituted censorship effectuated through extralegal sanctions. The doctrine drawn from the decision holds that when government suggestions are tantamount to commands and a private party’s compliance is not genuinely voluntary, the resulting suppression is attributable to the state. The line between permissible persuasion and unconstitutional coercion turns on whether the government’s communications carried a coercive threat the intermediary could not reasonably disregard.

Q: What are the limits of Packingham’s holding?

Packingham struck down a complete ban on social media access by a broad class of registrants, but it deliberately reserved several questions. The Court did not hold that registrants possess an affirmative right of access to social media in all circumstances. It did not address whether a more precisely drawn law, aimed at actual contact with minors, impersonation of children, or gathering information about minors, would survive scrutiny; Justice Kennedy’s opinion suggested such narrower measures remained available. It did not disturb individualized conditions of probation or supervised release, where a sentencing judge tailors limitations to a particular offender, nor did it address internet access within prisons and jails, where different doctrines apply. And because three justices joined the judgment without joining the majority’s reasoning, the opinion’s broadest language about cyberspace speaks for five justices rather than eight, which has led lower courts to apply it cautiously.

Q: Can a library patron challenge filtering under CIPA as applied?

Yes. Although United States v. American Library Association sustained the Children’s Internet Protection Act against a facial challenge, both the plurality and Justice Kennedy’s concurrence preserved as-applied challenges. The plurality credited the Solicitor General’s representation that libraries could readily disable blocking for adult patrons and unblock individual sites, adding that if experience proved otherwise, libraries and patrons could bring as-applied claims. Justice Kennedy made the judgment depend expressly on the premise that filtering could be disabled without significant delay, warning that a library refusing to lift blocking for an adult patron, or imposing burdensome procedures on such requests, could face a successful challenge. Justice Breyer likewise tied his concurrence to the ease of lifting the block. The facial victory for the statute therefore left open individual suits where the disabling mechanism failed in practice.

Q: What did Justice Thomas’s cert-denial statements about section 230 say?

Justice Thomas filed two statements respecting the denial of certiorari that invited the Court to reconsider the lower court consensus on section 230. In Malwarebytes, Inc. v. Enigma Software Group USA, LLC, 141 S. Ct. 13 (2020), he wrote at length that courts had discarded the common-law distinction between publisher and distributor liability, extended immunity to companies for their own content, and read subsection (c)(1) so capaciously that subsection (c)(2)’s narrower good-faith shield became surplusage. In Doe v. Facebook, Inc., 595 U.S. ___ (2022), he returned to the theme and again urged the Court to address the statute’s scope in an appropriate case. These statements have no precedential weight: a statement respecting a denial is not a holding, not even a dissent, and no colleague joined either one. Their significance was predictive and hortatory, and Gonzalez v. Google, which looked like the suitable case, returned the invitation unopened.

Q: What textual questions about section 230 remain open?

At least four puzzles await the Court’s first construction. First, whether subsection (c)(1), phrased as a rule of treatment rather than an immunity grant, does independent immunizing work beyond subsection (c)(2), or whether the Zeran line stretched a definitional clause so far that subsection (c)(2)’s good-faith limitation became surplusage. Second, whether algorithmic recommendation, which selects and amplifies already-published material without human review, counts as the traditional publisher function Zeran described. Third, whether the common-law distinction between publisher and distributor liability survives, so that a service with specific knowledge of unlawful material might face distributor-level responsibility. Fourth, where the boundary lies between an interactive computer service and an information content provider when a platform solicits submissions, structures their presentation, and promotes selected items. Each new product feature reopens these inquiries under a precedential regime the Court has never endorsed.

Q: Why does the First Amendment ceiling constrain statutory reform?

Because constitutional doctrine, not the liability statute, sets the outer boundary of what governments may require of platforms, and that boundary does not move when Congress rewrites the statute. Moody v. NetChoice held that compelled carriage on curated feeds must answer to the editorial-discretion precedents regardless of section 230’s fate, so a legislature that repeals the immunity cannot thereby license must-carry mandates. Reno and the COPA litigation established that criminal prohibitions on protected online expression face strict scrutiny and usually fail where less restrictive alternatives exist, a ceiling that applies to any successor statute Congress drafts. Even successful legislation, like CIPA, survived only by staying within the spending power. Reform efforts that aim at platform behavior must therefore be designed inside the First Amendment limits the Court has drawn, which is why debates over amending section 230 so often arrive at the same constitutional wall.