A technology lawyer receives two letters about the same video on the same platform. The first says the video defames a local business owner. The second says the video contains fifteen seconds of a copyrighted song. The platform leaves the first video up indefinitely and removes the second within hours. The client wants to know why the platform is allowed to do one and required to do the other, and whether the difference reflects company policy or law. The answer is law, and it is the comparison this article sets out: the communications liability rule that shields the platform without asking it to do anything, and the copyright rule that shields the platform only if it does a list of things. The two statutes are read about constantly and compared almost never, which is why working professionals misstate both. The full text treatment of the first statute is in the series article on section 230, and the full profile of the second is in the series article on the Digital Millennium Copyright Act. This article exists for the comparison itself: the nature of each protection, what each demands of a service, what notice does under each, what claims each covers, why two regimes were needed, and what follows for anyone who wants to reform them. A creator whose work is taken down, a startup counsel building a compliance plan, a journalist covering platform behavior, and a staffer drafting a bill all need the same map, because the asymmetry between the two regimes explains platform conduct more reliably than any theory about company incentives. The map starts with the two statutes stated on their own terms.

The immunity: 47 U.S.C. 230 as automatic protection

An immunity rather than a procedure

Section 230 of the Communications Act, codified at 47 U.S.C. 230, entered the statute books as section 509 of the Telecommunications Act of 1996, Public Law 104-104, signed on February 8, 1996, as part of that Act’s Communications Decency Act title. When the Supreme Court invalidated the Communications Decency Act’s indecency provisions in Reno v. ACLU, 521 U.S. 844 (June 26, 1997), it left section 230 intact. The provision has functioned ever since as the central federal shield for intermediaries that transmit, host, or display expression authored by others.

The provision’s defining architectural choice is that it grants an immunity, not a conditional harbor. The protection attaches automatically the moment its elements are met. It requires no registration with any agency, no designation of an agent to receive complaints, no adoption of a written policy, no takedown procedure, and no opt-in filing of any kind. A defendant qualifies by virtue of what it is and what the plaintiff’s claim asks a court to do, not by any compliance program it maintains. Congress has built other liability shelters on the opposite model, conditioning protection on a service’s satisfaction of enumerated prerequisites and its continued maintenance of them; the companion section of this article describes that conditional machinery. The contrast matters here only at the level of architecture. Section 230 asks nothing of the service and confers everything at the threshold.

Congress wrote the rationale into the statute itself. Section 230(b) declares it the policy of the United States, among other things, “to promote the continued development of the Internet and other interactive computer services and other interactive media,” and “to remove disincentives for the development and utilization of blocking and filtering technologies that empower parents to restrict their children’s access to objectionable or inappropriate online material.” The second of these policies responded to a concrete judicial development. In 1995 a New York trial court held an online service liable as the publisher of defamatory statements posted by third parties on its bulletin boards, treating the service’s exercise of editorial control, including its promulgation of content guidelines and its employment of moderators, as the basis for publisher status. The sponsors of the amendment, Representatives Christopher Cox and Ron Wyden, drafted the provision to reverse the resulting incentive: a service that screened, edited, or removed objectionable material would no longer risk greater liability for having moderated than for having done nothing at all.

Because the immunity turns on the character of the claim rather than on disputed questions of conduct, courts have frequently resolved it at the pleadings stage. Where a complaint itself establishes that the defendant provides or uses an interactive computer service, that the claim seeks to treat the defendant as the publisher or speaker of the material at issue, and that the material was supplied by someone else, there is often nothing for discovery to add, and judges have dismissed such complaints or granted judgment on the pleadings on that basis alone. Early resolution of this kind is part of the design rather than an accident of procedure. An immunity that could be invoked only after a full trial on the reasonableness of a defendant’s moderation practices would deter the very screening activity Congress sought to encourage, since the cost of litigating to judgment would itself become the penalty for moderating.

Section 230(a) sets out congressional findings that frame the immunity’s breadth. Congress found that the Internet offered “a forum for a true diversity of political discourse, unique opportunities for cultural development, and myriad avenues for intellectual activity,” and that interactive computer services had “flourished, to the benefit of all Americans, with a minimum of government regulation.” These findings explain why the protection takes the form of an immunity from the claim itself rather than an excuse to be proved at trial. An affirmative defense concedes the claim’s legal validity and offers a justification; the immunity of subsection (c)(1) denies the claim’s premise at the threshold, instructing courts not to treat the defendant as the publisher in the first place. The difference matters for litigation posture: because the statute negates an element of the plaintiff’s theory rather than supplying the defendant with an excuse, defendants may raise it in a motion to dismiss and need not await summary judgment, a procedural advantage that magnifies the substantive one.

The twenty-six words of subsection (c)(1)

The operative heart of the statute is a single twenty-six-word sentence in subsection (c)(1): “No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider.” Every element of the sentence does work. “No provider or user” extends the protection beyond companies to individuals, a point developed below. “Shall be treated as” frames the command as a rule of attribution rather than a standard of care; the question is never whether the defendant behaved reasonably, but whether the claim requires the court to regard the defendant as the author of someone else’s expression. “Publisher or speaker” captures the two roles through which liability for the content of expression ordinarily attaches. And “information provided by another information content provider” marks the boundary of the shield: the immunity covers third-party content, and it does not cover content the defendant itself created or developed.

The statute defines its central terms in subsection (f). An “interactive computer service” means “any information service, system, or access software provider that provides or enables computer access by multiple users to a computer server, including specifically a service or system that provides access to the Internet and such systems operated or services offered by libraries or educational institutions.” An “information content provider” means “any person or entity that is responsible, in whole or in part, for the creation or development of information provided through the Internet or any other interactive computer service.” The two definitions interlock. Content arrives through an interactive computer service from an information content provider, and the service that merely carries, hosts, or displays it is not to be treated as its publisher or speaker.

Courts applying the sentence have distilled it into a three-element inquiry. First, the defendant must be a provider or user of an interactive computer service. Second, the plaintiff’s claim must seek to treat the defendant as the publisher or speaker of the challenged information. Third, the information must have been provided by another information content provider. All three elements must be satisfied for the immunity to attach, and the failure of any one of them defeats it. The test’s second element does the heaviest lifting in contested cases, because plaintiffs often frame claims in terms of negligence, misrepresentation, or statutory duties precisely to avoid the language of publication. Courts look past the label to the substance: if the duty the plaintiff invokes would require the defendant to monitor, screen, or remove third-party content, the claim treats the defendant as a publisher regardless of the cause of action named in the complaint.

The logic of the sentence forecloses a characteristic family of claims. Defamation is the paradigmatic example: a plaintiff who sues a website for a false and damaging statement posted by a user asks the court to treat the website as the publisher of the statement, which is precisely what the sentence forbids. The same logic reaches claims framed to avoid the word “publisher.” A negligence claim alleging that a service failed to screen postings adequately, or failed to remove a posting promptly after receiving a complaint, still requires the court to treat the service as responsible for the publication; appellate courts have held such artful pleading ineffective, as the Zeran discussion below explains. The sentence does not, however, protect a defendant against claims that do not depend on publisher or speaker status. Courts have allowed claims to proceed where liability turns on a distinct promise or duty independent of publication, as in Barnes v. Yahoo!, 570 F.3d 1096 (9th Cir. 2009), while policing the line between a genuine independent obligation and a publisher-liability claim dressed in contract language.

A further consequence of the sentence’s structure is that the immunity is categorical rather than calibrated. The text contains no reasonableness standard, no knowledge qualifier, and no good-faith requirement. A service that never reviews its users’ postings receives the same protection as one that reviews them all; a service that leaves reported content in place receives the same protection as one that removes it within the hour. Congress made a deliberate trade. By detaching protection from any judgment about the defendant’s conduct, the statute avoids turning every lawsuit into an inquest on moderation practices and preserves the breathing room the policy section describes.

The Good Samaritan clause: subsection (c)(2)

Subsection (c)(1) protects a defendant against liability for third-party content. Subsection (c)(2), the Good Samaritan provision, protects a different act: the voluntary restriction of content. It provides that no provider or user of an interactive computer service shall be held liable on account of “any action voluntarily taken in good faith to restrict access to or availability of material that the provider or user considers to be obscene, lewd, lascivious, filthy, excessively violent, harassing, or otherwise objectionable, whether or not such material is constitutionally protected.” A second prong extends the same protection to “any action taken to enable or make available to information content providers or others the technical means to restrict access to material described in paragraph (1),” covering filtering software and parental-control tools.

The two subsections differ in both trigger and condition. Subsection (c)(1) is triggered by a claim about content the defendant did not create, and it imposes no condition on the defendant’s behavior; even a defendant that moderates nothing, or moderates badly, keeps the shield. Subsection (c)(2) is triggered by the defendant’s own act of restricting access, and it conditions protection on good faith. The distinction reflects the two hazards Congress perceived. Without (c)(1), a service faced liability for carrying the speech of others. Without (c)(2), a service faced liability for removing it: a user whose posting was deleted, or whose access was restricted, might sue the service for the removal, and the service that filtered aggressively might be treated as having assumed responsibility for everything it failed to filter. The Good Samaritan clause answers the second hazard by protecting good-faith efforts to keep objectionable material away from users, including efforts that prove mistaken about particular items.

The clause is notably broader than the indecency provisions that Reno struck down. By protecting good-faith restriction of material the provider “considers to be” within the enumerated categories, including material that is “otherwise objectionable” and “whether or not such material is constitutionally protected,” the clause leaves the judgment about objectionability to the provider rather than to a court. A service that removes lawful expression it deems objectionable may invoke the clause, provided the removal was voluntary and in good faith. The subjective framing matters: the question is what the provider considered the material to be, not whether a judge would agree with the characterization. The second prong extends the same logic to the tools trade, protecting the makers and distributors of filtering technologies against claims arising from the tools’ use.

The relationship between moderation and (c)(1) immunity deserves emphasis because the pre-statute case law had fused them. Under the approach the statute displaced, a service that exercised editorial control risked being deemed a publisher precisely because it had moderated. Section 230 severs that link. Appellate courts have held that ordinary editorial functions, selecting, arranging, editing, or deleting third-party material, do not convert a service into the creator or developer of the material and therefore do not forfeit (c)(1) protection. In Batzel v. Smith, 333 F.3d 1018 (9th Cir. 2003), for example, the operator of a listserv who received an allegedly defamatory email and published it to subscribers, making minor editorial alterations, retained the immunity; the court treated the modest editing as consistent with the role of a service rather than as authorship. The boundary, as the Roommates.com discussion below explains, lies at material contribution to the content’s creation or development, not at the exercise of editorial judgment.

What the immunity does not cover: subsection (e)

Subsection (e) lists five categories that the immunity leaves untouched, and the list defines the provision’s perimeter as precisely as (c)(1) defines its core. The exclusions operate as rules of construction: nothing in section 230 shall be read to impair or limit the listed bodies of law.

First, subsection (e)(1) preserves federal criminal enforcement. It provides that nothing in the section shall be construed to impair the enforcement of the federal provisions governing obscene or harassing telephone communications, the provisions on sexual exploitation of children, or “any other Federal criminal statute.” A service that violates federal criminal law cannot invoke the immunity as a shield against prosecution.

Second, subsection (e)(2) preserves intellectual property law, providing that nothing in the section shall be construed to limit or expand any law pertaining to intellectual property. Copyright and trademark claims therefore proceed outside section 230 entirely, under their own statutory frameworks. The practical consequence is a division of labor: expression-based tort claims against intermediaries meet the automatic immunity of section 230, while infringement claims meet the conditional machinery that Congress built separately for that purpose.

The phrasing of subsection (e)(2) is symmetrical: the section neither limits nor expands any law pertaining to intellectual property. The symmetry means that section 230 cannot be invoked to defeat an infringement claim, and equally cannot be invoked to create one. A copyright holder suing a service for hosting infringing material proceeds under the copyright statutes alone, unaffected by the communications provision in either direction. The division reflects a legislative judgment that the distinct policy balances of copyright, with its own notice-and-response machinery, should not be disturbed by a statute drafted for defamation and indecency.

Third, subsection (e)(3) addresses state law with a preemption rule rather than a blanket exclusion. It provides that nothing in the section prevents a State from enforcing any state law consistent with the section, and that no cause of action may be brought and no liability imposed under any state or local law inconsistent with it. State-law claims that depend on treating the defendant as the publisher or speaker of third-party content, such as state defamation or negligence claims, are therefore preempted; state enforcement that does not conflict with the immunity survives.

The preemption rule of subsection (e)(3) has done substantial work in practice because most claims against intermediaries arise under state law. Defamation, negligence, intentional infliction of emotional distress, and unfair-practice claims are creatures of state common law and state statutes, and each is preempted to the extent it requires treating the defendant as the publisher or speaker of third-party content. The provision’s second sentence operates as express preemption: it directs that no liability may be imposed under inconsistent state or local law, leaving courts to ask only whether the state claim conflicts with the federal immunity. Where the state claim can be adjudicated without attributing authorship to the defendant, the state law is consistent with the section and survives.

Fourth, subsection (e)(4) preserves communications privacy law, providing that nothing in the section limits the application of the Electronic Communications Privacy Act of 1986, its amendments, or section 705 of the Communications Act of 1934. Claims concerning the interception or disclosure of communications proceed under those statutes unaffected by the immunity.

Fifth, subsection (e)(5), added in 2018, excludes sex-trafficking-related claims meeting specified criteria. It provides that nothing in the section, other than subsection (c)(2)(A), shall be construed to impair or limit a federal civil action under 18 U.S.C. 1595 where the conduct underlying the claim constitutes a violation of 18 U.S.C. 1591; a state criminal prosecution where the conduct underlying the charge would constitute a violation of 18 U.S.C. 1591; or a state criminal prosecution where the conduct underlying the charge would constitute a violation of 18 U.S.C. 2421A and the promotion or facilitation of prostitution is illegal in the jurisdiction where the defendant’s conduct was targeted. The parenthetical preserving (c)(2)(A) is deliberate: even within the trafficking exclusion, the Good Samaritan protection for good-faith blocking remains intact.

Zeran v. AOL: the end of distributor liability

The foundational appellate construction of the immunity came within two years of the statute’s enactment. In Zeran v. AOL, 129 F.3d 327 (4th Cir. Nov. 12, 1997), the Fourth Circuit confronted the argument that section 230 left a gap for distributor liability, and closed it.

The facts were stark. In the aftermath of the Oklahoma City bombing, an unidentified person posted messages on an America Online bulletin board advertising “Naughty Oklahoma T-Shirts” with slogans celebrating the bombing, listing the home telephone number of Kenneth Zeran, a Seattle resident with no connection to the postings. Zeran began receiving threatening calls. He notified AOL, which removed the postings, but new postings kept appearing, including one directing callers to ask for Ken. Zeran sued, alleging that AOL had been negligent in failing to remove the defamatory material promptly after receiving notice, failing to screen for similar postings, and failing to post retractions.

The Fourth Circuit affirmed judgment on the pleadings for AOL. Its central holding was that section 230 bars distributor liability as well as publisher liability. At common law, distributors such as bookstores and libraries faced liability for defamatory material only if they knew or had reason to know of the defamation, a more forgiving standard than the strict liability imposed on publishers. Zeran argued that this distinction survived the statute: once AOL received notice of the false postings, it became a distributor with knowledge and could be held liable for failing to act. The court rejected the argument on textual and structural grounds. Distributor liability, the court reasoned, is merely a subset of publisher liability; both theories treat the defendant as responsible for the publication of another’s statement. Because subsection (c)(1) forbids treating a service as the publisher or speaker of third-party content, it forecloses the distributor theory along with the publisher theory.

The consequence is that actual notice of unlawful content does not defeat the immunity. A service that learns that a posting is defamatory, and leaves it in place, remains protected. The court grounded this result in the statute’s evident purpose. Notice-based liability would confront every service with an impossible choice upon receiving a complaint: remove the challenged material immediately, without any ability to verify the complaint’s accuracy, or face potential liability for leaving it up. Complaints are cheap to file and costly to investigate, and the rational response to such a regime would be deletion on demand, which would chill lawful speech and punish the service for the volume of complaints it receives rather than for any wrongdoing. Congress, the court concluded, had chosen to eliminate that regime root and branch rather than to calibrate it.

The artful-pleading problem that Zeran addressed has recurred in new factual settings, and courts have applied the same analysis. In Doe v. MySpace, Inc., 528 F.3d 413 (5th Cir. 2008), the plaintiffs sued a social-networking service for negligence after a minor was assaulted by someone she met through the service, alleging that the service had failed to implement age-verification safeguards. The Fifth Circuit held the claim barred: although framed as negligence rather than defamation, the claim sought to hold the service liable for its role in publishing users’ communications, and the failure-to-protect theory was, in substance, an attempt to treat the service as the publisher of third-party content. The decision illustrates how the second element of the (c)(1) inquiry operates. A plaintiff cannot evade the immunity by renaming the claim; the court asks what the claim requires the defendant to have done differently, and if the answer is to have screened, monitored, or removed the communications of others, the claim is a publisher claim.

Zeran also confirmed that the immunity does not depend on the service’s passivity. AOL exercised editorial functions, and the court held that irrelevant; the statute protects the service regardless of the editorial role it assumes, because the question under (c)(1) is who supplied the content, not how actively the service managed its forums. The decision’s procedural posture reinforced the point made earlier about early resolution: the court decided the immunity question on the pleadings, without discovery into AOL’s moderation practices, because the complaint itself showed that the claim treated AOL as the publisher of content supplied by someone else.

The Zeran construction has been widely adopted by the regional circuits. Its notice rule is the reason the immunity functions as an automatic protection rather than a qualified one: if knowledge stripped the shield, the shield would protect only the ignorant, and every plaintiff would plead notice.

Roommates.com: the material-contribution boundary

If Zeran defines the immunity’s breadth, Fair Housing Council of San Fernando Valley v. Roommates.com, LLC, 521 F.3d 1157 (9th Cir. Apr. 3, 2008) (en banc), defines its boundary. A service that crosses from carrying content to creating or developing it becomes an information content provider as to that content, and the immunity does not attach.

Roommates.com operated a website that matched individuals seeking shared housing. To use the service, subscribers created profiles by answering a questionnaire. The site required answers to questions concerning sex, sexual orientation, and whether the subscriber would bring children, and it supplied pre-populated answers through dropdown menus from which subscribers selected. Fair housing organizations sued, alleging that the site’s questions and answer choices facilitated housing discrimination in violation of the Fair Housing Act. The Ninth Circuit, sitting en banc, held that the website was “responsible, in whole or in part, for the creation or development” of the resulting profiles within the meaning of subsection (f)(3), and therefore was an information content provider as to that information rather than a mere interactive computer service carrying it.

The court’s reasoning turned on the site’s role in shaping the content. Roommates.com had not simply provided blank fields into which users typed whatever they wished. It had designed the questions, required answers as a condition of participation, and channeled responses into pre-written categories that embodied the allegedly discriminatory criteria. By soliciting the information and structuring its expression, the site materially contributed to the content’s development. The court distinguished this conduct from the provision of neutral tools. A search engine that indexes the web, or a payment processor that handles transactions, supplies functionality without shaping the substance of any particular communication; such tools do not make the provider responsible for the content they carry. Roommates.com, by contrast, had become a co-developer of the profiles because its questionnaire design determined what the profiles said.

The opinion drew the boundary with care by holding that the site retained immunity for a different portion of the same website. Subscribers could also add free-form essays in an “Additional Comments” section, and as to those open-ended entries the site was not an information content provider; the immunity applied. The result illustrates that the boundary is content-specific rather than service-wide. A single website can be a protected intermediary as to material its users author independently and an unprotected content provider as to material it helps create, with each item of content assessed on its own terms.

The Ninth Circuit had reached the opposite result on similar technology five years earlier, and the contrast sharpens the boundary. In Carafano v. Metrosplash.com, Inc., 339 F.3d 1119 (9th Cir. 2003), a dating service provided questionnaires through which users built profiles, including multiple-choice questions with pre-written answers, and an unknown person used the tools to post a defamatory profile of an actress. The court held the service immune: although it supplied the questionnaire, the defamatory content itself came from the user, and providing the structure through which content is expressed did not make the service the content’s creator. Roommates.com distinguished that situation on the ground that its questionnaires dictated discriminatory substance as a condition of participation. Together the two decisions locate the line: supplying neutral structure is carrying, while designing the questions and compelling answers that embody the alleged wrong is developing.

The “material contribution” formulation has since served as the standard shorthand for the inquiry. The question is not whether the service touched the content, since editing, formatting, and selecting do not forfeit protection, but whether the service’s contribution shaped the content in a way connected to its alleged unlawfulness. A service that requires users to input information through structured prompts of its own design, and then publishes the resulting profiles as its product, has crossed the line the statute draws in subsection (f)(3).

The user as a protected party

Subsection (c)(1) protects “provider or user,” and the inclusion of users extends the immunity to individuals who republish the expression of others. A person who forwards an email to a mailing list, reposts another’s review, or shares another’s message is a user of an interactive computer service, and a claim seeking to hold that person liable as the publisher of the forwarded content meets the same statutory bar that protects the platform. The logic is identical: the claim asks the court to treat the forwarder as the publisher or speaker of information provided by another information content provider, which the twenty-six words forbid.

Batzel v. Smith illustrates the principle at the individual level. The defendant operated a listserv devoted to a local controversy; he received an email containing allegedly defamatory accusations and published it to the list’s subscribers, making minor editorial changes. The Ninth Circuit held the immunity applicable, treating the defendant as a user and provider who had not become the creator of the content despite his modest alterations. The decision confirms that ordinary acts of republication, selecting an item for forwarding and making incidental edits in the course of transmission, do not convert the republisher into the author.

The protection for users carries its own boundary, parallel to the one Roommates.com drew for services. A user who repeats another’s statement verbatim remains within the immunity; a user who adds original defamatory assertions of their own becomes an information content provider as to the added material. Courts assess the added portion separately, so that the shield covers the republished content while leaving the republisher answerable for what they themselves contributed.

The user protection has grown in practical significance as republication became a central mode of online expression. Sharing, reposting, and forwarding functions allow a single item of third-party content to propagate across networks in seconds, and each intermediary recipient is a user of an interactive computer service as to the content they pass along. The immunity means that the chain of republication does not multiply publisher liability with each link; liability follows authorship, not transmission. The result is symmetrical with the treatment of services: just as a platform is not the publisher of its users’ postings, a user is not the publisher of the postings they relay, and in both cases the boundary is drawn at original contribution.

FOSTA: the single enacted narrowing

Since its enactment in 1996, Congress has narrowed the immunity’s scope by statute exactly once. The Allow States and Victims to Fight Online Sex Trafficking Act of 2017, enacted as Public Law 115-164 on April 11, 2018, responded to appellate decisions that had applied the immunity to dismiss civil trafficking claims against online classified-advertising services. The statute added subsection (e)(5), described above, carving sex-trafficking claims meeting its criteria out of the immunity’s reach, while expressly preserving the (c)(2)(A) Good Samaritan protection for good-faith blocking.

FOSTA made two companion changes outside section 230. It created a new federal offense at 18 U.S.C. 2421A, punishing anyone who owns, manages, or operates an interactive computer service with the intent to promote or facilitate the prostitution of another person, and it amended the definition of “participation in a venture” in 18 U.S.C. 1591 to reach those who knowingly assist, support, or facilitate sex trafficking. The three provisions work together: the new and amended criminal provisions define the underlying conduct, and subsection (e)(5) ensures that the immunity does not block the specified civil and criminal actions founded on that conduct.

The carve-out is narrow by design. It does not withdraw the immunity from all claims touching prostitution or trafficking; it withdraws it only from the specified federal and state actions where the conduct underlying the claim constitutes or would constitute a violation of the enumerated federal offenses. A civil claim against a service that lacks the required connection to section 1591 or section 2421A conduct continues to meet the full immunity. The precision reflects the provision’s history: Congress wrote the immunity in categorical terms in 1996, and when it later decided that a particular category of wrongdoing required a different balance, it legislated the exception with equal precision rather than qualifying the immunity generally.

The singularity of FOSTA is itself informative about the statute’s design. Proposals to condition, qualify, or repeal the immunity have been introduced and debated, but only the trafficking carve-out has been enacted into law. The immunity that Congress created in 1996 therefore operates in essentially the form Zeran construed: automatic, without procedural prerequisites, surviving notice, bounded only by the authorship line that Roommates.com drew and the five exclusions that subsection (e) enumerates.

What must a service file or adopt before section 230 protects it?

Nothing. Section 230 requires no registration with any agency, no designated agent for complaints, no written policy, no takedown procedure, and no filing of any kind. A defendant qualifies by virtue of what it is and what the plaintiff’s claim asks the court to do, not by any compliance program it maintains.

Into what three elements have courts distilled the twenty-six words?

Courts ask three questions. The defendant must be a provider or user of an interactive computer service. The claim must seek to treat the defendant as the publisher or speaker of the challenged information. And the information must have been provided by another information content provider. Failure of any element defeats the immunity.

What does subsection (e)(2) say about intellectual property, exactly?

Subsection (e)(2) is symmetrical: nothing in section 230 shall be construed to limit or expand any law pertaining to intellectual property. The provision cannot be invoked to defeat an infringement claim, and it cannot be invoked to create one. Infringement proceeds under the copyright and trademark statutes alone.

How does section 230 treat a service’s ordinary editing of user posts?

Ordinary editorial work does not forfeit the immunity. In Batzel v. Smith (2003), a listserv operator who published a defamatory email with minor editorial alterations retained protection, because selecting, arranging, editing, or deleting third-party material is consistent with the role of a service. The boundary lies at material contribution to the content’s creation.

Which state laws survive section 230 under subsection (e)(3)?

Subsection (e)(3) preserves state law that is consistent with the section, and bars any cause of action or liability under state or local law that is inconsistent with it. State defamation and negligence claims that require treating a service as the publisher of third-party content are preempted; state enforcement that can proceed without that attribution survives.

The conditional harbor: 17 U.S.C. 512 and its compliance checklist

The Digital Millennium Copyright Act, Public Law 105-304, signed October 28, 1998, implemented two World Intellectual Property Organization treaties adopted December 20, 1996, but the statute reached further than treaty obligation required. Its most consequential addition to domestic copyright law was not the anticircumvention prohibition at 17 U.S.C. 1201 but the limitation on liability created at 17 U.S.C. 512. Where the anticircumvention rules gave copyright owners a new weapon against the circumvention of technological protection measures, section 512 gave online intermediaries a shield: a set of safe harbors that cap exposure to monetary relief for copyright infringement occurring through their systems, provided the provider satisfies a detailed compliance checklist. The safe harbors were the congressional answer to the litigation anxiety of the mid-1990s, when district courts produced conflicting answers about whether an Internet service provider or bulletin board operator could be held liable for infringing copies that users placed on its servers. Congress chose a compromise. It did not declare intermediaries immune. It constructed a conditional harbor, one that rewards specific conduct, including prompt response to notices, termination of repeat offenders, and accommodation of standard technical measures, while withdrawing protection from providers that benefit from infringement they could control or that ignore what they plainly know.

The relationship between this regime and 47 U.S.C. 230 must be understood first, because confusion between the two statutes produces the most common errors in commentary about platform liability. Section 230(e)(2) expressly provides that nothing in section 230 limits or expands any law pertaining to intellectual property. Copyright claims were therefore carved out of section 230 immunity from the beginning. A provider sued for copyright infringement cannot invoke section 230 at all. The copyright regime at section 512 stands on its own, with its own definitions, its own conditions, and its own procedures. This structural choice reflected a deliberate legislative judgment: the balance between encouraging investment in communications infrastructure and protecting creators was too delicate to be folded into a general immunity, so Congress wrote a separate, detailed statute for it. The practical consequence is that a single platform can be immune from a defamation claim about user content under section 230 and simultaneously face full exposure for a copyright claim about the same material if it fails the section 512 checklist.

Section 512 organizes its protection into four safe harbors, each keyed to a different technical function of the service provider. The statute defines service provider broadly, covering both entities that offer transmission, routing, or connections for digital online communications and providers of online services or network access, so the harbors reach everyone from backbone carriers to consumer-facing platforms. The first harbor, section 512(a), addresses transitory digital network communications: the mere conduit function. It protects providers that transmit, route, or provide connections for material through a system or network controlled or operated by the provider, provided the transmission was initiated by someone other than the provider, the process is carried out through an automatic technical operation without selection of the material, the provider does not select the recipients, no copy accessible to anyone but the intended recipients is maintained on the system longer than reasonably necessary, and the material is transmitted without modification of its content. Classic Internet service providers carrying packets between endpoints are the paradigm case, and the harbor reflects the judgment that a carrier which merely moves bits at a user’s direction should not answer for what those bits contain.

The second harbor, section 512(b), covers system caching: the intermediate and temporary storage of material on a system or network, done to make onward transmission more efficient. Caching was, and remains, essential to the functioning of the Internet, because storing frequently requested content closer to the user reduces bandwidth costs and latency. The harbor conditions require that the material be made available online by a person other than the provider, transmitted to the subsequent user at that person’s direction, and stored through an automatic technical process. The provider must comply with rules concerning the refreshing, reloading, or other updating of the material when specified by the source, must not interfere with standard technology used at the originating site to obtain information about the material, and, upon proper notice, must remove or disable access to cached material that has been removed or disabled at the originating site. The caching harbor thus preserves the efficiency of the network while preventing stale or infringing copies from lingering once the original has been taken down.

The third harbor, section 512(c), is the one that dominates litigation and commercial practice. It covers information residing on systems or networks at the direction of a user: the storage function that every social media platform, video sharing site, marketplace, cloud storage service, and web host performs. Because user-uploaded content is the lifeblood of the modern platform economy, section 512(c) is the provision that determines whether services such as YouTube, eBay, or a commercial web host face potentially ruinous statutory damages for infringements committed by their users. Its conditions are the most elaborate in the statute and the subject of the most contested case law, and they are examined in detail below.

The fourth harbor, section 512(d), covers information location tools: referring or linking users to an online location containing infringing material or activity, through directories, indexes, references, pointers, or hypertext links. Search engines are the paradigm beneficiaries. Its conditions mirror those of section 512(c), adapted to the linking context, including the requirements of no disqualifying knowledge, expeditious removal or disabling of access upon proper notice, and the absence of direct financial benefit coupled with the right and ability to control the infringing activity.

Before reaching the activity-specific conditions of any harbor, every provider must satisfy the threshold requirements of section 512(i), which apply across all four harbors. A provider that fails these threshold conditions cannot invoke any safe harbor, regardless of how scrupulously it otherwise behaves. The first threshold condition requires the provider to have adopted and reasonably implemented a policy that provides for the termination in appropriate circumstances of subscribers and account holders who are repeat infringers. The second requires the provider to accommodate and not interfere with standard technical measures, defined as measures developed by broad consensus of copyright owners and service providers in an open, fair, voluntary, multi-industry process, available on nondiscriminatory terms, and not imposing substantial costs or burdens on providers. The repeat infringer policy requirement was for years treated as a formality, with providers maintaining policies on paper while doing little to enforce them. That era ended with the Fourth Circuit’s decision in the Cox Communications litigation, discussed below, which made clear that a policy is not reasonably implemented when the provider systematically undermines it.

The first prong of section 512(i) also contains a concrete administrative obligation that is frequently overlooked in summaries of the statute: to be eligible for the limitations on liability, a service provider must have designated an agent to receive notifications of claimed infringement, and the designation must be made available to the public on the provider’s website and filed with the Copyright Office. Congress originally required designations on paper; the Copyright Office later established an electronic registration system and required providers to re-register in the new system by the end of 2017, invalidating paper designations that were not renewed. A provider that neglects this registration cannot claim the harbor’s protection no matter how quickly it responds to notices in practice, because the statutory eligibility turns on the formal designation. Counsel advising platforms therefore treat agent registration and maintenance as a first-order compliance item, alongside the substantive conditions.

The full section 512(c) checklist builds on the threshold conditions. A provider of user-directed storage qualifies for the harbor only if three additional requirements are met. First, the provider must not have actual knowledge that the material or an activity using the material on the system or network is infringing. Second, in the absence of such actual knowledge, the provider must not be aware of facts or circumstances from which infringing activity is apparent: the so-called red flag standard. Third, upon obtaining such knowledge or awareness, the provider must act expeditiously to remove, or disable access to, the material. The statute also addresses knowledge obtained through notification: if the provider receives a notification of claimed infringement that complies with the statute’s formal requirements, described below, and responds expeditiously by removing or disabling access to the material claimed to be infringing, the provider is deemed to satisfy the knowledge and red flag conditions for that material. A fourth substantive condition applies to the financial dimension: the provider must not receive a financial benefit directly attributable to the infringing activity in a case in which the provider has the right and ability to control such activity. The financial benefit prong borrows from the common law doctrine of vicarious copyright liability, which holds one party liable for another’s infringement when the first has the right and ability to supervise the infringing activity and a direct financial interest in it. The harbor thus denies protection to providers that profit from infringement they could have prevented, even as it protects providers whose revenue bears no direct relationship to specific infringements.

The red flag standard deserves careful attention because it is the provision most often misread. Awareness of facts or circumstances from which infringing activity is apparent is an objective test: it asks what a reasonable person would have understood from the information the provider actually possessed, not what the provider subjectively believed. Congress nevertheless set the bar high. The legislative history indicates that red flags must be brightly red indeed, obvious to a reasonable person, before the awareness prong is triggered. A general awareness that a service is used for some infringing activity does not disqualify a provider; the awareness must relate to specific instances of infringement. This distinction, between general and specific knowledge, became the central battleground in the Viacom litigation against YouTube, and the Second Circuit’s resolution of it continues to shape platform compliance programs.

The notice-and-takedown mechanics of section 512(c)(3) translate these knowledge standards into an operational procedure. A notification of claimed infringement is effective only if it contains the elements the statute lists: a physical or electronic signature of a person authorized to act on behalf of the owner of an exclusive right allegedly infringed; identification of the copyrighted work claimed to have been infringed, or, if multiple works at a single site are covered by one notification, a representative list; identification of the material claimed to be infringing, with information reasonably sufficient to permit the provider to locate it; information reasonably sufficient to permit the provider to contact the complaining party; a statement that the complaining party has a good faith belief that the use is not authorized by the copyright owner, its agent, or the law; and a statement, under penalty of perjury, that the information in the notification is accurate and that the complaining party is authorized to act on behalf of the owner. Notifications that substantially comply are effective; minor technical defects do not automatically invalidate them, but the statute instructs providers and courts to disregard elements that do not substantially comply when evaluating the notice. In practice, the volume of notices is enormous, and large platforms process them through automated systems, a fact that has generated its own controversies about accuracy and over-removal.

The counter-notice and put-back procedure at section 512(g) gives the user whose material was removed a path to contest the takedown. A subscriber who believes material was removed or disabled as a result of mistake or misidentification may send the provider a counter-notification containing the subscriber’s signature, identification of the removed material and its prior location, a statement under penalty of perjury that the subscriber has a good faith belief the material was removed as a result of mistake or misidentification, and the subscriber’s name, address, telephone number, and consent to the jurisdiction of the federal court for the district of the subscriber’s address, or, if outside the United States, for any judicial district in which the provider may be found, plus an agreement to accept service of process from the complaining party or its agent. Upon receiving an effective counter-notice, the provider must promptly provide the original complainant with a copy and inform the complainant that the provider will replace the removed material or cease disabling access to it in ten to fourteen business days. If the complainant does not file a court action seeking to restrain the subscriber within that window, the provider must restore the material. The put-back mechanism is the statute’s principal safeguard against erroneous or abusive takedowns, and a provider that follows the procedure in good faith is shielded from liability to the subscriber for the removal, while a provider that restores material after a proper counter-notice faces no liability to the original complainant for the restoration.

Section 512(f) polices abuse of the notice system itself. Any person who knowingly materially misrepresents that material or activity is infringing, or that material or activity was removed or disabled by mistake or misidentification, is liable for damages, including costs and attorneys’ fees, incurred by the alleged infringer, the copyright owner, or the service provider as a result of the provider’s reliance on the misrepresentation in removing or disabling access or in replacing the material. The misrepresentation must be knowing and material; negligent or mistaken notices do not trigger the provision. The scope of the good faith belief statement in a takedown notice was clarified by the Ninth Circuit in Lenz v. Universal Music Corp., 801 F.3d 1126 (9th Cir. 2015). Stephanie Lenz had posted a short home video of her children dancing to Prince’s “Let’s Go Crazy,” and Universal Music had issued a takedown notice; Lenz sued under section 512(f), arguing that Universal could not have formed a good faith belief of infringement without considering whether her use was fair use. The Ninth Circuit agreed that fair use is not merely an affirmative defense but a use authorized by law, so a copyright holder must consider fair use before sending a takedown notice. The court stopped short of demanding a full legal analysis in every case; it held that the consideration of fair use must be subjective and in good faith, and that the formation of a good faith belief does not require investigation beyond what the circumstances reasonably call for. The decision nevertheless imposed a meaningful discipline on automated notice programs that fire without any human or algorithmic assessment of obvious fair uses, and it remains the leading statement of the duty that notice senders owe under section 512(f).

The litigation record under section 512(c) is dominated by two appellate decisions that together define the compliance boundary. In Viacom International, Inc. v. YouTube, Inc., 676 F.3d 19 (2d Cir. 2012), a group of copyright owners sought more than a billion dollars in damages for alleged infringement of thousands of works on YouTube. The Second Circuit affirmed summary judgment for YouTube on the central knowledge question, holding that the actual knowledge and red flag provisions of section 512(c) require knowledge of specific and identifiable instances of infringement, not merely a general awareness that the platform hosts infringing material. Evidence that YouTube’s founders knew infringement was rampant on the service could not, standing alone, disqualify the provider from the harbor. The court also addressed willful blindness, holding that the doctrine could in principle establish the knowledge or awareness required by the statute, but that the plaintiffs had not shown YouTube deliberately avoided learning about specific infringements. The decision further clarified the right and ability to control prong, holding that it requires something more than the mere ability to remove or block access to materials posted on a service; it demands the kind of substantial influence over users’ activities that the common law vicarious liability cases require, such as prescreening content or directing users’ infringing choices. Viacom thus entrenched a provider-friendly reading of the knowledge provisions while leaving the door open for cases built on item-specific evidence.

The Fourth Circuit’s decision in BMG Rights Management (US) LLC v. Cox Communications, Inc., 881 F.3d 293 (4th Cir. 2018), supplied the counterweight, and it is the most important appellate loss for a provider under the statute. BMG, acting through its enforcement agent, sent Cox hundreds of thousands of notices identifying subscribers engaged in peer-to-peer infringement of BMG’s music catalog. The evidence showed that Cox maintained a repeat infringer policy on paper but systematically declined to implement it: the company capped the number of infringement notices its abuse department would process, blacklisted the enforcement agent’s notices so they never reached subscribers, and reactivated terminated accounts, all to avoid losing paying customers. A jury found Cox contributorily liable and denied the safe harbor, awarding substantial damages, and the Fourth Circuit affirmed the denial of the harbor. The court held that section 512(i)’s requirement of a reasonably implemented repeat infringer policy means what it says: a provider that deliberately renders its own policy toothless has not implemented it, reasonably or otherwise, and forfeits the harbor’s protection. Cox therefore lost the shield not because of what it knew about any particular file, but because of a threshold failure that infected its entire compliance posture. The decision reverberated through the industry, prompting providers to audit their graduated response systems and to document termination decisions, because the harbor’s threshold conditions were no longer a formality that diligent-looking paperwork could satisfy.

These decisions together map the compliance checklist that sophisticated providers now follow. The designated agent registration must be current with the Copyright Office and mirrored on the provider’s site. The repeat infringer policy must exist in writing, must define repeat infringement with sufficient clarity, must be applied in practice with records of warnings and terminations, and must not be sabotaged by internal caps, blacklists, or revenue-driven exceptions. Takedown operations must remove or disable access expeditiously upon receipt of compliant notices, with expeditiousness measured against the circumstances and the volume at issue. Content review and trust-and-safety teams must be trained to recognize that red flags, while rare, can arise from internal communications or specific complaints that identify infringing material with precision. Notice-sending operations, whether human or automated, must incorporate the Lenz discipline of considering fair use before asserting infringement. And the business model must be examined for direct financial benefit tied to infringing activity that the provider has the right and ability to control, because that prong operates independently of the knowledge provisions.

The final structural point is the one most often lost in public discussion of the statute. Losing the safe harbor does not create liability. The harbors are limitations on liability, affirmative shields that a qualifying provider may raise. A provider that fails the checklist simply stands where it would have stood had section 512 never been enacted: exposed to claims of direct, contributory, or vicarious infringement decided under ordinary copyright principles. The copyright owner must still prove its case, including ownership of a valid copyright, copying of protected expression, and the elements of the applicable secondary liability theory. This asymmetry explains why the litigation stakes are so high and why compliance programs are so elaborate: the difference between the harbor and the common law baseline is the difference between capped, predictable exposure and the full range of copyright remedies, including statutory damages that can reach six figures per work infringed. Congress built a conditional harbor rather than an immunity precisely to keep that pressure in place, using the threat of ordinary liability to purchase the cooperation of intermediaries in the enforcement system the statute created.

Three related provisions of the section 512 framework complete the picture of the bargain Congress struck. Section 512(m) addresses the monitoring question directly: safe harbor eligibility shall not be conditioned on a service provider monitoring its service or affirmatively seeking facts indicating infringing activity, except to the extent consistent with a standard technical measure. This is the statutory foundation for the principle that providers need not police their networks proactively. The knowledge provisions of section 512(c) are triggered by what the provider actually knows or by red flags apparent from information it actually possesses, not by what diligent monitoring might have uncovered. The no-duty-to-monitor rule explains why courts have rejected arguments that a provider forfeits the harbor by failing to deploy filtering technology on its own initiative: Congress considered imposing such a duty and declined, leaving filtering to voluntary cooperation and to the standard technical measures process. The provision has its limits. It does not protect a provider that deliberately structures its operations to avoid learning about infringement, which is the territory of willful blindness, and it does not excuse a provider from acting on the knowledge it does acquire. But as a default allocation of responsibility, it places the burden of identifying infringement on the copyright owner, which is the economic logic underlying the entire notice-and-takedown system.

Section 512(h) provides copyright owners with an expedited subpoena mechanism to learn the identity of an alleged infringer from the service provider, issued by the clerk of a federal court upon a proper notification and a sworn declaration, without the filing of a lawsuit. The provision was designed for the era of peer-to-peer file sharing, when copyright owners could see an IP address engaged in infringement but could not name the subscriber behind it. Its reach was narrowed by the D.C. Circuit in Recording Industry Association of America, Inc. v. Verizon Internet Services, Inc., 351 F.3d 1229 (D.C. Cir. 2003), which held that the subpoena power extends only to providers storing material at the direction of users under section 512(c), not to mere conduits under section 512(a) that simply transmit data. The practical effect was to push copyright owners seeking subscriber identities toward the ordinary discovery process in filed litigation, often under the standards for anonymous speaker identification developed in later cases. Section 512(j), in turn, limits the injunctive relief available against a qualifying provider, restricting courts to orders restraining the provider from providing access to infringing material at identified locations, ordering termination of specified subscriber accounts, and other relief the court considers necessary to prevent or restrain infringement of a specified work at a specified location, with special solicitude for the burden on the provider’s network and the least burdensome effective remedy. Together these provisions show a statute calibrated at every joint: the shield is real, but it is fenced with conditions, and the enforcement tools given to copyright owners are likewise fenced with limits.

The expeditiousness requirement of section 512(c)(1)(A)(iii) has generated its own body of interpretation, because the statute supplies no fixed deadline. Courts evaluate expeditiousness in light of the circumstances, considering the volume of notices, the provider’s resources and systems, and the nature of the material at issue. Takedowns completed within a day or two of a compliant notice are routinely treated as expeditious, while delays stretching into weeks without justification invite findings of noncompliance. The standard is deliberately flexible, and that flexibility cuts both ways. Large platforms with automated notice intake can remove material within hours, and copyright owners have come to expect that speed; smaller providers with manual review processes are judged against their own capabilities rather than against the industry leaders. What the standard does not tolerate is inertia: a provider that sits on compliant notices, or that routes them into a queue where they languish while the infringing material continues to draw traffic, cannot claim the harbor’s protection for the delay period. Compliance programs therefore treat notice response time as a monitored metric, with escalation procedures for backlogs and documented timestamps establishing when each notice arrived and when the corresponding material came down.

The financial benefit prong of section 512(c)(1)(B) imports the common law of vicarious copyright liability into the harbor analysis. At common law, a defendant is vicariously liable for another’s infringement when the defendant has the right and ability to supervise the infringing activity and a direct financial interest in its exploitation, as developed in cases such as Shapiro, Bernstein and Co. v. H.L. Green Co., 316 F.2d 304 (2d Cir. 1963), and Fonovisa, Inc. v. Cherry Auction, Inc., 76 F.3d 259 (9th Cir. 1996). The harbor denies protection where both elements coincide: the provider receives a financial benefit directly attributable to the infringing activity, and the provider has the right and ability to control it. The directness qualifier matters. Advertising revenue generated across a platform as a whole, without a nexus to specific infringing material, has generally not satisfied the provision; courts look for a tighter connection, such as revenue sharing tied to the infringing content itself. The right and ability to control element, as clarified in Viacom, requires more than the capacity to remove material after the fact; it demands the kind of practical control over users’ conduct that the vicarious liability cases contemplated, such as prescreening uploads, directing the selection of content, or exercising editorial authority over what users post. The ordinary platform, which offers storage to all comers on equal terms and removes material upon notice, typically satisfies this prong. The provision bites where the business model is built around infringement that the provider could curtail, which is exactly the category of enterprise Congress meant to exclude from the harbor.

The aftermath of the two landmark decisions illustrates how the statute’s conditions interact in practice. After the Second Circuit’s Viacom ruling, the case returned to the district court, which again granted summary judgment to YouTube on remand, and the parties settled in 2014 without further appellate review, leaving the Second Circuit’s framework as the governing statement of the knowledge standard in that circuit and a persuasive authority elsewhere. Copyright owners adjusted their strategies accordingly, shifting resources from broad showings of platform-wide infringement toward item-specific evidence: internal emails identifying particular clips, deposition testimony about particular uploads, and notice histories tied to particular works. The Lenz litigation followed a similar arc of clarification followed by settlement. After the Ninth Circuit established that fair use must be considered before a takedown notice issues, the parties settled, and the doctrinal contribution of the case became the standard itself rather than any damages award. Section 512(f) claims remain difficult to win, because the statute demands knowing and material misrepresentation, and courts have held that the good faith belief inquiry is subjective: a notice sender that honestly, even if mistakenly, believes infringement has occurred does not misrepresent. The provision’s practical function is therefore less as a damages remedy than as a disciplinary boundary that keeps the notice system honest, with Lenz supplying the fair use component of that honesty.

The administrative side of compliance has grown more demanding over the statute’s life. The designated agent registration that section 512(c)(2) contemplates was for years a paper filing mailed to the Copyright Office, and many providers let their designations go stale without consequence. The Copyright Office replaced the paper system with an electronic directory, required all providers to re-register electronically by the end of 2017, and invalidated designations that were not renewed. Providers must now maintain their listings, update them when contact information changes, and pay the prescribed fee, with registrations expiring after three years if not renewed. A lapsed registration is a threshold defect that can forfeit the harbor regardless of the provider’s substantive conduct, and defense counsel in infringement litigation now routinely check the directory as a first step. The standard technical measures prong of section 512(i) has seen less development, because no identification technology has yet emerged through the broad multi-industry consensus process the statute describes. Voluntary filtering and fingerprinting systems exist, but they remain proprietary rather than consensus standards, so the accommodation duty has had limited practical bite. Its significance is structural: Congress built the statute to accommodate future consensus technologies without requiring new legislation, a design choice that has kept the framework adaptable even as the technologies for identifying infringing content have evolved.

The relationship between the harbors and the remedies available outside them closes the analysis. A qualifying provider is shielded from monetary relief, including actual damages, profits, and statutory damages, which under 17 U.S.C. 504(c) can reach one hundred fifty thousand dollars per work for willful infringement. That exposure is the engine of the entire compliance apparatus: the prospect of aggregated statutory damages across thousands of works is what makes the checklist worth following. The harbor does not, however, eliminate all judicial remedies. Section 512(j) preserves the availability of injunctive relief against qualifying providers, albeit in the narrowed forms the statute specifies. And the harbor’s protection extends only to copyright claims. A provider that satisfies every element of section 512(c) remains fully exposed to trademark, patent, trade secret, and other non-copyright claims arising from the same user content, subject to whatever defenses those bodies of law supply. This is the deeper meaning of the section 230(e)(2) carve-out with which this discussion began. Congress did not create a single unified immunity for online intermediaries. It created a general immunity for non-intellectual-property claims under section 230 and a separate, conditional, compliance-driven regime for copyright under section 512. The two statutes answer different questions, impose different obligations, and reflect different legislative bargains. Any comparison of the two regimes that treats them as variations on a single theme misses the architecture: one is a shield granted by status, the other a harbor earned by conduct, and the conduct it demands is specified in the checklist down to the registration of an agent, the termination of repeat infringers, and the expeditious removal of material upon proper notice.

Five axes, one verdict

Comparing 47 U.S.C. 230 with 17 U.S.C. 512 looks at first like comparing two tools built for different workshops, and in a sense it is: one shields platforms from liability for third party expression generally, the other shields them from liability for third party copying specifically. But the comparison rewards close attention because the two provisions share a skeleton. Both were Congress’s answer to the same structural problem, namely that internet intermediaries transmit far more material than any human workforce could review, and both answer it by conditioning or removing liability so that the network can function. The difference is not in the problem but in the philosophy of the fix. One provision grants immunity with no strings attached. The other grants a conditional harbor that is earned and kept only through ongoing compliance. Everything downstream flows from that design choice.

First axis: nature, unconditional immunity against conditional harbor

The communications provision operates automatically. A qualifying interactive computer service needs to do nothing to earn its protection: no registration, no designated agent, no posted policy, no response protocol. The statute’s operative sentence is self executing. A platform qualifies, the content came from someone else, and the claim treats the platform as the publisher of that content; when those three conditions hold, immunity follows without any further showing of care, diligence, or process. The absence of procedural prerequisites is not an oversight but the core feature. Congress wrote the provision to let services set their own moderation standards, high or low, without fear that a good faith mistake or a careless lapse would cost them the shield. The price of the immunity is not compliance; there is no compliance to buy it with. That is why a service that does nothing at all, that reviews nothing and moderates nothing, stands in exactly the same legal position as one with an elaborate trust and safety apparatus, at least as far as the statute is concerned.

The copyright provision is the mirror image. Section 512 offers no automatic protection. A service must qualify for the harbor by satisfying a checklist of conditions before any claim arises, and it must keep satisfying them continuously. It must designate an agent to receive notices of claimed infringement and register that agent with the Copyright Office. It must adopt and reasonably implement a policy for terminating the accounts of repeat infringers, and inform its users of that policy. It must respond expeditiously to proper notices by removing or disabling access to the identified material. It must lack actual knowledge of infringement, lack awareness of facts or circumstances from which infringing activity is apparent, and it must not receive a financial benefit directly attributable to the infringing activity where it has the right and ability to control that activity. Failure on any condition means the harbor is unavailable, and losing the harbor is not a finding of liability; it is the loss of a defense, returning the service to the ordinary rules of secondary copyright liability where plaintiffs must prove their case the hard way. The practical effect is a standing compliance burden: the statute assumes that earning protection for copying requires verifiable procedures, while the communications provision assumes that protection for expression requires none.

Second axis: trigger, what has to happen for protection to exist or disappear

Under the communications provision, nothing is required of the service. Immunity attaches by operation of law to any qualifying defendant in any covered case, and it cannot be lost through passivity, carelessness, or bad moderation judgment. The statute contains no conduct test and no revocation mechanism. A service that ignores complaints, enforces its rules erratically, or enforces no rules at all keeps the shield in full. This is the provision’s most controversial property and its most deliberate one: Congress wanted services free to moderate or not moderate without liability turning on how well they did the job.

Under the copyright provision, protection is contingent and revocable in practice. A service earns the harbor by building the compliance apparatus described above, and it loses the harbor, for the particular work at issue or more broadly, when it fails a condition. Miss the expeditious response deadline on a valid notice and the harbor is gone for that material. Tolerate known repeat infringers and the policy condition fails. Take a direct financial benefit from infringement one has the power to stop and the benefit condition fails. Each failure is discrete and each is curable going forward, but the structure means the service’s legal position depends at every moment on its own ongoing conduct. The contrast with the first axis sharpens here: one statute asks nothing and therefore cannot be failed; the other asks continuously and therefore can be.

One clarification matters for honest comparison. Losing the copyright harbor does not create liability by itself. It merely deprives the service of the fast exit from the lawsuit, forcing it to litigate under doctrines like contributory and vicarious infringement, where knowledge, control, and financial benefit become plaintiff’s burdens to prove. The communications immunity, by contrast, is a defense that ends the case at the threshold. The asymmetry is real: the copyright regime penalizes noncompliance by removing a shield, while the communications regime offers no comparable lever because it attaches no conditions to begin with.

Third axis: notice, the fork where the two doctrines diverge most visibly

Notice is the single sharpest point of contrast. Under the communications provision, actual knowledge of unlawful third party content does not defeat immunity. The Fourth Circuit said so directly in Zeran v. AOL, 129 F.3d 327 (4th Cir. 1997), rejecting the argument that notice of defamatory postings transformed a service into a publisher liable for them. The court reasoned that making liability turn on notice would create an impossible dilemma: a service receiving notices would have to investigate each claim’s truth, an undertaking at internet scale that no provider could perform, and the rational response would be to remove everything flagged, chilling speech far beyond what the law requires. Later courts have followed Zeran’s core holding even as they have refined its edges. The result is that a platform may be told, specifically and credibly, that material on its servers is defamatory or harassing, and it may leave the material in place indefinitely without incurring liability under the provision. Knowledge changes nothing.

Under the copyright provision, notice changes everything. A proper notice of claimed infringement, identifying the work and the location of the offending material, triggers an obligation to act expeditiously to remove or disable access. The statute even specifies the notice’s required elements and gives the service a counter notification process to restore material when the poster disputes the claim. Failure to respond expeditiously forfeits the harbor for the identified material. This is a genuine notice and takedown regime: rights holders police the network, services act as the enforcement arm, and the statute’s safe harbor is the consideration paid for that cooperation.

The policy logic behind each rule is internally coherent. Copyright notices are, at least in the statute’s conception, verifiable without judging truth: the notice identifies a registered work, the service compares the posted file to the work, and the match is a technical question. Defamation and harassment notices are not verifiable that way. Determining whether a statement is false, whether it concerns a public figure, whether privilege applies, whether context changes meaning, these are adjudicative judgments that courts themselves struggle with, and Congress decided not to deputize platforms as courts. Whether that judgment was correct is the subject of the reform debate, but the internal logic of each notice rule fits its subject matter.

The knowledge asymmetry: red flags against Zeran

Between the notice axis and the scope axis sits a related divergence worth naming separately: what each statute does with knowledge that falls short of a formal notice. The copyright provision disqualifies a service that is aware of facts or circumstances from which infringing activity is apparent, the so called red flag test. A service cannot keep the harbor by averting its eyes from the blatantly infringing. Courts have construed this narrowly, requiring that the infringement be obvious to a reasonable person without investigation, but the principle stands: willful blindness to the obvious forfeits protection. The doctrine reflects copyright’s verifiability premise. Some infringement announces itself, a full pirated film labeled as such on a file sharing index, and the statute expects the service to recognize what requires no adjudication.

The communications provision has no equivalent. Zeran’s rule is not limited to formal notices; it holds that knowledge of unlawful content, however acquired and however specific, does not convert the service into the publisher of that content. There is no red flag exception to the immunity, no duty triggered by the obviousness of the wrong. A service may host content whose unlawfulness is apparent on its face and remain fully protected, because the statute’s protection does not depend on the service’s mental state at all. This is the most uncompromising feature of the immunity and the one most alien to lawyers trained in tort or copyright, where knowledge almost always matters. Congress made knowledge irrelevant for speech claims for the reason the Fourth Circuit articulated: once knowledge triggers a duty, every complaint becomes a potential liability event, and the platform’s rational response is removal without inquiry. The copyright statute accepts a narrow version of that duty because the inquiry it demands is shallow. The communications statute rejects even the narrow version because the inquiry its subject matter demands is deep.

The asymmetry also shapes the reform proposals described below. Any proposal to add a knowledge based trigger to the communications provision, a duty to act on red flag unlawful content, must confront the reason the red flag test works in copyright and would strain elsewhere. In copyright, the flag is red because the match is technical. In defamation, harassment, or threats, the flag’s color depends on contested judgments about truth and context, which means the proposed duty would either be so narrow as to change little or so broad as to recreate the Zeran dilemma. The comparison does not prove such a duty unworkable. It identifies the precise point where the analogy to copyright stops carrying weight.

Fourth axis: scope, how far each shield reaches

The communications provision covers essentially the entire civil law of third party content, with defined carve outs. It bars claims arising from publication of third party material: defamation, negligence, false light, intentional infliction of emotional distress premised on publishing, unfair trade practices built on hosting someone else’s claims, and similar theories. The carve outs are specific: intellectual property claims are excluded, federal criminal liability is untouched, communications privacy statutes including the Electronic Communications Privacy Act are preserved, and specified sex trafficking claims under the regime created by FOSTA, P.L. 115-164, enacted April 11, 2018, survive. That list is worth reading carefully because it reveals congressional intent. Congress did not leave intellectual property in the immunity by accident; it wrote the exclusion so that copyright and trademark would be governed by their own regimes. The trafficking carve out, added more than two decades after enactment, remains the only enacted substantive narrowing of the immunity’s reach.

The copyright provision covers copyright alone. It offers no protection against defamation claims, no protection against privacy claims, no protection against any non copyright theory. A service that satisfies every Section 512 condition to the letter may still face a defamation suit over the same hosted material, and in that suit the copyright compliance counts for nothing. Conversely, a service fully immunized under the communications provision against a defamation claim over hosted material enjoys no copyright protection from that same immunity; the intellectual property exclusion sends the copyright question to Section 512, where the service must meet the checklist or defend on ordinary infringement grounds.

This division of labor is the point of the fifth axis.

Fifth axis: why both exist, complementarity by design

The two statutes are not competitors. They are a deliberate division of the intermediary liability problem. When Congress enacted the communications provision as section 509 of the Telecommunications Act of 1996, P.L. 104-104, signed February 8, 1996, it excluded intellectual property from the immunity’s scope. Two years later, Congress enacted the Digital Millennium Copyright Act, P.L. 105-304, signed October 28, 1998, and gave copyright its own intermediary regime in 17 U.S.C. 512. The exclusion in the first statute is precisely the reason the second statute exists. Congress decided that expression and copying raised different enough problems to warrant different architectures: a no conditions immunity for the open ended, hard to verify world of speech claims, and a conditional harbor for the comparatively verifiable world of copyright claims.

That design choice explains features of each statute that look like bugs when viewed in isolation. The communications provision’s tolerance of known unlawful content looks like indifference until one remembers that Congress reserved judgment intensive claims from any notice based duty. The copyright provision’s bureaucratic checklist looks like overkill until one remembers that Congress needed a system rights holders could actually use at scale. Each is fitted to its domain. The friction arises only when a single piece of content implicates both domains at once, which the modern internet does constantly.

The practical consequence: same post, two regimes

Consider a single user post that is both allegedly defamatory and allegedly infringing, a copied news photograph accompanied by a false accusation about a private individual. Under the communications provision, the service may leave the post up indefinitely after receiving detailed notice of the defamation, with no legal exposure, because notice does not defeat the immunity and no duty to investigate attaches. Under the copyright provision, the service must act expeditiously on a proper notice identifying the photograph or lose its harbor and face ordinary secondary liability. The service’s obligations turn entirely on the legal theory named in the complaint. The same file, the same server, the same platform conduct produces opposite duties depending on whether the plaintiff frames the grievance as a speech tort or an infringement claim. There is nothing paradoxical about this once the design is understood; it is the complementarity working as built. But it is the fact pattern that makes the comparison concrete, and it is the fact pattern that drives the reform debate, because reformers look at the defamation side of that example and ask why the copyright side’s machinery cannot be adapted to it.

The burden ledger: what compliance actually costs

The five axes describe legal structure. The burden ledger describes what that structure costs to live inside, and the ledger deepens the contrast. Under the communications provision, the compliance cost of immunity is zero by design. A startup with two founders and no legal department enjoys the same threshold defense as the largest platform on earth, and the defense is typically won on a motion to dismiss, before discovery, on the pleadings. That early exit is part of the design’s economic logic: Congress wanted the immunity to protect not only against liability but against the cost of defending, because litigation expense alone can determine what speech infrastructure survives. The Zeran line of cases delivers that by making the three part test answerable from the complaint itself: the defendant is a service, the content came from a third party, the claim treats the service as publisher. When those appear on the face of the pleading, the case ends.

Under the copyright provision, the compliance cost is a standing line item, and it scales with volume. Designating an agent is trivial. Building and staffing a notice processing operation that can respond expeditiously to millions of notices is not. Maintaining a repeat infringer policy that is actually implemented, keeping records sufficient to show reasonable implementation, training reviewers to distinguish proper notices from defective ones, operating a counter notification and putback workflow with its ten to fourteen business day clock, these are institutional commitments that persist whether or not any lawsuit is filed. Large services have built automated matching systems to absorb the volume, and those systems themselves become part of the compliance story. Small services face the same statutory checklist with a fraction of the resources, which is one reason the copyright regime’s burdens fall unevenly across the market. The comparison matters for reform because every proposal to add procedure to the communications provision is implicitly a proposal to add this kind of standing cost to every service the procedure covers, and the history of Section 512 suggests the cost will be borne most easily by incumbents with existing compliance departments. A reform that conditions immunity on process tends, all else equal, to entrench the platforms large enough to afford the process.

There is a second, subtler cost. The copyright checklist gives plaintiffs and their counsel a discovery roadmap. Every condition is a potential deposition topic: show us the repeat infringer policy, show us how it was implemented, show us the log of notices and response times, show us the financial benefit analysis. Litigation under Section 512 is fact intensive in a way that litigation under the communications provision is not, and that fact intensity is itself a policy choice. Congress accepted expensive, granular litigation over copying because the underlying rights are federal, registered, and economically measurable. It chose cheap, threshold litigation over speech claims because the underlying questions are constitutional in flavor and resistant to quick resolution. Any proposal to move the communications provision toward the copyright model should be understood as a proposal to move from the first kind of litigation to the second.

FOSTA and what the single carve out teaches

The trafficking carve out added by FOSTA deserves separate attention because it is the only enacted instance of Congress narrowing the communications immunity, and therefore the only empirical test of what narrowing does. The statute, P.L. 115-164, enacted April 11, 2018, created a federal criminal offense for owning or operating a service with intent to promote or facilitate prostitution and carved specified trafficking related civil claims out of the immunity. Supporters argued that the blanket shield had protected platforms that knowingly facilitated exploitation, and that a targeted exception would preserve the immunity’s core while addressing its worst failure. Critics predicted that the exception’s breadth and the severity of the new criminal exposure would cause platforms to over remove lawful sexual expression, harm reduction resources, and consensual adult content far beyond the trafficking the statute targeted.

Both predictions found support in what followed, which is precisely why the episode is instructive for the broader comparison. Services did remove large volumes of content and shutter entire categories of classifieds and personals, including material with no connection to trafficking, because the legal risk calculus favored aggressive removal once immunity was no longer certain. At the same time, advocates for trafficking victims reported continued exploitation migrating to less visible venues. The lesson is not that the carve out was right or wrong; reasonable observers disagree, and the empirical record remains contested. The lesson is structural and directly relevant to the five axes. When Congress moved one category of claims from the unconditional column to a conditional one, the market response was exactly what the Zeran court predicted in the notice context: uncertainty about the boundary produced removal well beyond the boundary. FOSTA is the closest thing the communications regime has to a natural experiment in adding conditions, and its results counsel humility about how precisely any new procedural regime can be calibrated.

The verdict

The comparison is the reform debate. That is the verdict, and the deciding factor is structural: nearly every serious proposal to change the communications provision amounts to a proposal to make it resemble the copyright provision by adding notice requirements, procedural duties, or conditional protection. Proposals described by their mechanics rather than their sponsors fall into recognizable families. One family would impose a notice and response obligation for specified categories of unlawful content, requiring services to act within defined timeframes after receiving complaints meeting statutory criteria, with immunity retained only for services that comply. Another family would replace blanket immunity with a reasonable care duty, under which services that take commercially reasonable steps to prevent foreseeable harm keep protection while those that do not face ordinary tort standards. A third family would tier obligations by service size, reach, or function, imposing heavier procedural burdens on the largest platforms while leaving smaller services under something closer to the current rule. Each family borrows, consciously or not, from the copyright statute’s architecture: define the trigger, specify the procedure, condition the protection on compliance. The copyright harbor is the working model of what a conditional intermediary regime looks like in American law, so every debate about conditioning the communications immunity is a debate conducted in Section 512’s shadow. One makes no claim here about which outcome is preferable. The point is narrower and firmer: the five axes above are not academic taxonomy but the actual terrain on which reform proposals stand or fall, because each proposal must answer each axis. What conditions, if any, should protection require. What conduct triggers or defeats it. What role notice plays. How far the shield reaches. And whether the two regimes should converge or remain complementary.

Each proposal must also specify the enforcement mechanism, and here the copyright statute offers two distinct templates. One template conditions a defense: comply with the procedure and keep the shield, fail and litigate on the merits. The other template creates direct liability for procedural failure, fining or penalizing the service for the missed deadline or the absent policy regardless of the underlying content’s lawfulness. Section 512 uses the first template; losing the harbor removes a defense rather than creating liability. Reform proposals for the communications provision vary on this point, and the choice matters enormously. A condition on the defense preserves the background tort law as the ultimate arbiter, which means the hardest questions of truth, privilege, and context still get decided by courts. A direct procedural penalty bypasses those questions entirely, punishing the process failure as such. The copyright comparison illuminates the difference because Section 512’s restraint on this point, it never fines a service for slow takedown as an independent wrong, is part of why the regime has remained politically durable.

A further question every proposal must answer is the analogue of the designated agent and the registered work. The copyright system functions because notices arrive in a standardized form, directed to a known recipient, referencing an identifiable work with a registration behind it. A notice regime for defamation or harassment has no equivalent anchor. There is no registry of true statements, no filing office for reputation, no technical match to perform. Proposals that borrow the notice mechanic must therefore invent the verification infrastructure from scratch: who certifies that a complaint is proper, what elements must it contain, who judges sufficiency, and what happens to defective or abusive notices. The copyright statute answers these questions with specificity, down to the elements of a valid notice and the perjury statement. A communications provision analogue would need its own equally detailed answers, and the difficulty of drafting them is itself evidence for the verdict. The comparison is the reform debate not only because reformers borrow from Section 512 but because the borrowing exposes exactly where the analogy breaks.

The counter reading: why notice and takedown may not transfer

The most common assumption in the reform debate deserves direct scrutiny: the idea that a notice and takedown model, proven in copyright, would straightforwardly improve the communications regime. The comparison itself supplies the strongest reasons for doubt, and they deserve equal care with the accountability concerns that motivate reform.

Start with verifiability. A copyright notice is, in the typical case, machine checkable. The notice identifies a registered work, points to a file, and the service can compare the two without resolving contested questions of fact. Automated systems perform this matching at enormous scale, and while the automation generates its own documented problems, the underlying comparison is a technical operation. A defamation notice is not checkable that way. Verifying a defamatory claim requires judging whether the statement is true or false, whether the subject is a public or private figure, whether a privilege such as fair report applies, whether opinion or hyperbole defeats the claim, and whether surrounding context changes meaning. Harassment and threats claims require similar contextual judgment. These are determinations that courts make through adversarial process with rules of evidence, and the Zeran court’s warning was that imposing them on platforms at scale would produce not careful adjudication but indiscriminate removal. A service facing liability for leaving content up and no liability for taking it down will take it down; the incentive structure of any notice based regime points one way.

The copyright system provides the empirical basis for that warning. Its documented problems are well known to scholars and practitioners who study the regime: over removal driven by risk aversion, in which services remove material on thin or defective notices rather than test the claim; automated abuse, in which bad actors file fraudulent or mistaken notices to suppress competitors, critics, or lawful uses; and particular harm to fair use and other lawful expression that lacks a well resourced defender to file counter notices. These are not speculative pathologies. They are the observed output of a notice and takedown system operating for more than two decades at internet scale, under conditions more favorable than any speech tort regime would enjoy, because copyright at least offers the technical matching that defamation lacks. If a verifiable subject matter produces this much over removal, a non verifiable subject matter can be expected to produce more.

None of this dismisses the accountability concerns on the other side. The current rule permits services to ignore even well documented unlawful content with complete legal safety, and critics argue with force that this produces real, unremedied harm to defamation victims, harassment targets, and others who cannot practicably sue the original speaker. Proponents of reform further argue that large modern platforms bear little resemblance to the fragile startups the 1996 Congress imagined, and that entities with sophisticated moderation infrastructure can bear procedural duties the early internet could not. Those arguments go to institutional capacity and moral responsibility, and they are not answered by pointing to copyright’s flaws. But copyright’s flaws are the best available evidence of what the proposed machinery does in practice, and any honest reform debate must weigh the documented costs of notice based systems alongside the documented costs of the current immunity. The comparison cuts both ways by design: Section 512 shows that conditional protection is workable, and it also shows what conditional protection costs. A debate that cites only one of those lessons is not using the comparison. It is using half of it.

The deeper frame is institutional: the question is not whether unlawful content should have consequences but who decides, under what procedures, with what error costs. The current communications regime assigns the hard speech questions to courts, where plaintiffs must identify the actual speaker and prove their case under substantive law, and it accepts that some meritorious claims will go unremedied because the speaker is anonymous, judgment proof, or beyond reach. The copyright regime assigns the initial decision to the service, acting on a rights holder’s say so, with courts available later through counter notification and litigation, and it accepts that some lawful expression will be suppressed in the interim because speed favors the complainant. Each regime chooses its characteristic error. The immunity errs toward leaving harmful content up. The harbor errs toward taking lawful content down. Reform proposals that would graft notice duties onto the communications provision are, at bottom, proposals to change which error the system prefers, and the honest version of that debate names the tradeoff instead of promising that procedure eliminates it.

This is why the judgment here stays restricted to institutional design. Reasonable people can hold that the current allocation of error is indefensible, that platforms with vast resources should not be permitted to ignore documented harm, and that Congress should condition the immunity on reasonable care. Reasonable people can equally hold that deputizing private platforms as adjudicators of truth and context will suppress dissent, criticism, and minority expression at far greater scale than any court based system, and that the documented record of the copyright regime is a preview, not a coincidence. Both positions take the comparison seriously. The first reads Section 512 as proof that conditional regimes can function. The second reads it as proof of what they cost. Presenting the accountability concern means acknowledging that real victims face real barriers under the current rule, including the practical impossibility of suing anonymous speakers and the inadequacy of platform goodwill as a remedy. Presenting the over removal concern means acknowledging that the copyright system’s error rate is not a design flaw to be engineered away but a structural feature of any system that rewards removal and punishes retention. Equal care is not false balance; it is the recognition that the five axes describe a genuine dilemma rather than a puzzle with a clean solution.

The verdict stands on that ground. The comparison between the two statutes is the substance of the reform debate because every serious proposal is an argument about which features of the copyright regime the communications regime should adopt: its conditions, its notice mechanics, its procedural checklist, or none of them. Understanding the axes, the burden ledger, and the single enacted carve out does not dictate an answer. It disciplines the question, which is the most any comparison can honestly do.

The Two-Regime Comparison Table

Dimension	47 U.S.C. 230	17 U.S.C. 512
Nature of protection	Immunity from suit on publisher or speaker theories, attaching automatically	Safe harbor limiting liability, conditioned on compliance
Conditions to qualify	None; no procedural prerequisites	Designated agent, repeat infringer policy, no disqualifying knowledge, no barred financial benefit, expeditious response to notices
Effect of notice	None; protection survives actual knowledge of unlawful content	Proper notice triggers a duty of expeditious removal; failure forfeits the harbor
Claims covered	Nearly all civil claims arising from third-party content, except intellectual property, federal criminal law, communications privacy, and specified trafficking claims	Copyright claims only
What a service must do	Nothing required; protection exists from the first day of operation	Register an agent, answer notices, maintain policies, and avoid disqualifying conduct
Losing protection	Not available as a design feature; applies by statute unless an exclusion applies	Failing any condition removes the defence; liability is then decided on ordinary copyright law
Underlying theory	Knowledge and process are kept out of liability so services neither fear hosting nor fear moderating	Knowledge and process are the point, giving rights holders a workable enforcement channel

The Closing Assessment

The comparison yields a single governing fact, and it deserves a name. Same post, two regimes: the legal obligations attached to a piece of user content are determined by the theory of liability in the complaint, not by the content itself or by the harm it causes. A defamatory falsehood about a named person can sit on a platform forever with no legal exposure for the host. The same host, holding the same file, faces a structured compliance obligation the moment a copyright notice names it. This is not a bug in either statute and not a contradiction between them. The communications immunity was drafted to keep knowledge and process out of the liability question entirely, so that services would not fear either leaving content up or taking it down. The copyright harbor was drafted to put knowledge and process at the center, so that rights holders would get a workable enforcement channel without litigating every infringement. Each statute solves the problem it was given with the instrument suited to that problem, and the mismatch that frustrates observers is the point at which the two problems meet in one file.

The verdict follows from the comparison rather than from any preference. The deciding factor is this: most serious proposals to reform the communications immunity are, at bottom, proposals to make it resemble the copyright harbor by adding notice and procedure. Conditioning protection on reasonable content moderation, requiring response to complaints within a set window, and grading obligations by service size are all architectures borrowed from the conditional model. The comparison is therefore not academic background for the reform debate but the reform debate’s actual substance, and a reader who understands the five axes can read any proposal’s mechanics and say exactly which features it imports and which it does not. The counterweight also follows from the comparison. Copyright notices can be checked against a registered work by automated systems, while defamation and harassment claims demand judgments about truth, privilege, context, and intent that no machine makes reliably. The documented over-removal, automated abuse, and repeat infringer disputes in the copyright system are the best available evidence of what a notice-driven regime produces at scale, and any proposal that imports that machinery should be read with that evidence in mind rather than with the assumption that procedure improves outcomes by default.

The study path depends on the reader. A student should begin with the immunity, because its logic is the simpler of the two and because every reform argument presupposes it; the decision order that works is the 1996 provision, then the 1998 statute, then this comparison, and the series guide on studying US legislation carries that order for the whole field. A staffer drafting or evaluating a reform bill should read the five axes before reading any proposal, so that the mechanics of what is being imported are visible on first reading. A researcher tracing the doctrinal development should hold the Zeran notice holding and the Viacom knowledge standard as the two anchor points, since they are the decisions that fixed what notice means under each regime. A journalist covering platform conduct should start from the practical consequence rather than from company statements: when a platform leaves material up or takes it down, the question to ask is which statute’s obligations govern the complaint, because that answer predicts the conduct more reliably than any account of incentives. The one enacted narrowing of the immunity, the 2018 FOSTA amendment, is traced in the series article on FOSTA and SESA, and the claims about what repealing the immunity would do are answered in the series article on section 230 myths. Readers who want to work the comparison against their own cases can keep their notes in the VaultBook legislation study notebook, and students preparing coursework on the material can use the ReportMedic civics study tool to turn the five axes into a study set.

Frequently Asked Questions

Q: What is the difference between section 230 and the Digital Millennium Copyright Act?

Section 230 of the Communications Act, codified at 47 U.S.C. 230, is an immunity that prevents a provider or user of an interactive computer service from being treated as the publisher or speaker of third-party content. It attaches automatically, with no conditions to satisfy and no procedures to follow. The Digital Millennium Copyright Act, at 17 U.S.C. 512, is a conditional safe harbor for online service providers facing copyright claims. A service earns it only by registering a designated agent for infringement notices, responding expeditiously to proper notices, maintaining a reasonably implemented policy for terminating repeat infringers, and avoiding disqualifying knowledge or financial benefit tied to infringement it can control. Section 230 expressly excludes intellectual property claims, which is precisely why copyright needed its own regime. The practical difference is stark: a platform may leave an allegedly defamatory post up indefinitely with no legal exposure, yet must act promptly on a copyright notice or risk its defence.

Q: Why does the Digital Millennium Copyright Act exist if section 230 already protects platforms?

Section 230 contains an explicit exclusion for intellectual property law, found at 47 U.S.C. 230(e)(2), which states that nothing in the section limits or expands any law pertaining to intellectual property. Courts have read that exclusion to cover copyright and trademark claims, so a platform facing a copyright suit cannot invoke section 230 as a defence. When Congress drafted the Digital Millennium Copyright Act two years later, it was legislating into the gap that exclusion created. The DMCA’s safe harbors at 17 U.S.C. 512 give online services a structured set of defences for copyright claims, conditioned on compliance with notice-and-takedown and related obligations. The two statutes are therefore complementary by design rather than redundant. Section 230 handles nearly everything except intellectual property, and the DMCA handles copyright on its own conditional terms.

Q: Must a platform act on a Digital Millennium Copyright Act notice but not a defamation complaint?

Yes, and the asymmetry is the single most useful fact in this comparison. Under section 230, a platform has no legal obligation to act on a defamation complaint, a harassment report, or any other notice of allegedly unlawful non-copyright content. The Fourth Circuit held in Zeran v. AOL that the immunity bars distributor liability as well as publisher liability, so knowledge of the content does not create a duty to remove it. Under the DMCA’s section 512(c), by contrast, a service must respond expeditiously to a proper takedown notice to keep its safe harbor for the stored material at issue. Failing to act does not automatically create liability, but it forfeits the defence, leaving the service to face ordinary copyright claims. The same platform, holding the same post, therefore faces entirely different obligations depending on the legal theory in the complaint.

Q: Is section 230 an immunity while the Digital Millennium Copyright Act is a safe harbor?

That distinction is exactly right and it is the one most summaries get wrong. Section 230 is an immunity in the full sense: it operates at the threshold of a lawsuit, barring claims that depend on treating the defendant as the publisher or speaker of third-party content, and it imposes no procedural prerequisites, no registration, no notice response, and no compliance checklist. The DMCA’s section 512 creates safe harbors, which are conditional defences. A service qualifies for them only by satisfying a list of requirements, and each of the four harbors carries its own conditions rather than sharing a single standard. The terminology is not interchangeable. An immunity protects without being earned; a safe harbor protects only while its conditions are maintained. Confusing the two is the source of the common error that a platform must respond to complaints generally, when only the copyright regime imposes that kind of duty.

Q: Does section 230 cover copyright claims under the Digital Millennium Copyright Act?

No. Section 230(e)(2) expressly preserves all laws pertaining to intellectual property, and courts have read “intellectual property” to encompass copyright and trademark claims. A platform sued for hosting infringing music, video, or images cannot invoke section 230. It must look to the DMCA’s safe harbors at 17 U.S.C. 512, which were written specifically for copyright disputes. The exclusion is also why the two regimes coexist without conflict: Congress fenced copyright out of the immunity and then built a separate conditional regime for it. A service facing both a defamation claim and a copyright claim over the same post will defend the first under section 230 and the second under the DMCA, applying entirely different rules to the two defences. The boundary is statutory and deliberate, not a gap in either law.

Q: Which is harder to lose, section 230 protection or Digital Millennium Copyright Act safe harbor?

The section 230 immunity is far harder to lose, because there is effectively no mechanism for losing it. It attaches by operation of statute to qualifying content and claims, and it cannot be forfeited by a service’s conduct short of falling into one of the statutory exclusions. A platform that ignores every complaint it receives keeps the immunity intact. The DMCA safe harbor is lost by failing any one of its conditions: an unregistered designated agent, a missing or unimplemented repeat infringer policy, disqualifying knowledge of infringement, or a financial benefit directly attributable to infringing activity the service has the right and ability to control. The major internet provider that lost safe harbor protection in the Cox litigation lost it because its repeat infringer policy existed on paper but was not reasonably implemented. The contrast captures the whole comparison: one regime protects regardless of behavior, the other protects only through behavior.

Q: Do reform proposals want to make section 230 more like the Digital Millennium Copyright Act?

In substance, many of them do. The most serious reform proposals share a recognizable architecture: conditioning the immunity on reasonable content moderation practices, requiring services to respond to complaints within defined windows, grading obligations by the size or reach of the service, or removing protection for categories of claims unless specified procedures are followed. Those are the mechanics of the DMCA safe harbor transplanted into the communications immunity. The deciding factor in this article’s verdict is that the reform debate is, at its core, an argument about importing one statute’s architecture into the other. Recognizing this lets a reader evaluate any proposal on its actual terms rather than on rhetoric about immunity. It also frames the honest objection: a notice-driven regime carries the documented over-removal and automated-abuse problems of the copyright system, so the reform question is whether those costs are worth paying for non-copyright harms.

Q: Which should a student study first, section 230 or the Digital Millennium Copyright Act?

Section 230 comes first, because it is the simpler regime and because the DMCA’s design only makes sense against it. A student who begins with the 1996 provision learns the baseline rule: no publisher or speaker treatment for third-party content, no duties, no procedures. The 1998 statute then appears as the structured exception, the conditional regime Congress built for the one major category of claims the baseline excludes. Studying them in that order also matches the reform debate, where proposals are evaluated by what they would add to the unconditional baseline. The recommended sequence is the immunity’s text and its key cases, then the four safe harbors and their conditions, then a direct comparison across nature, trigger, notice, scope, and purpose. Students who reverse the order tend to treat notice-and-takedown as the general rule of platform liability, which is exactly backwards.

Q: Can a platform lose DMCA safe harbor protection while keeping section 230 immunity?

Yes, and this happens routinely. The two protections attach to different claims under different conditions, so a service can forfeit one while the other remains fully intact. A platform that fails to register its designated agent or neglects its repeat infringer policy loses the DMCA safe harbor for the copyright claim, but section 230 continues to bar any defamation, negligence, or housing discrimination claim arising from the same content, because the immunity has no behavioral conditions to forfeit. Conversely, a service that falls within one of section 230’s exclusions, such as a federal criminal prosecution, still has no section 230 defence but may retain its DMCA harbor for parallel copyright claims. The regimes are independent tracks. This is why compliance planning treats them as separate checklists rather than a single program, and why losing a copyright defence does not expose a service to the non-copyright claims it feared most.

Q: Why is a copyright notice easier to verify than a defamation complaint?

A copyright notice can be checked against a fixed, registered work. The takedown sender identifies a copyrighted recording, film, or image, and the platform can compare the posted file against it, often with automated matching systems. The verification is technical and bounded: is this the work, and was it posted without authorization. A defamation complaint requires judgments no machine makes reliably. The recipient must determine whether the statement is false, whether it concerns the complainant, whether it is privileged as opinion or fair comment, whether the subject is a public figure, and what the surrounding context communicates. Those are legal conclusions about meaning, and reasonable people disagree about them constantly. This asymmetry is the strongest argument against simply importing notice-and-takedown into defamation and harassment: the copyright system works because the underlying claim is verifiable, and the documented over-removal in that system is itself a warning about what a similar regime would produce for harder judgments.

Q: Does the DMCA safe harbor apply to trademark or defamation claims?

No. The safe harbors at 17 U.S.C. 512 apply to copyright claims only. A service facing a trademark suit over counterfeit listings or a defamation suit over a user review cannot invoke the DMCA’s notice-and-takedown framework or its liability limitations. Trademark claims are litigated on trademark principles, and defamation claims against a service run into section 230, which will bar them unless the service itself created or developed the content. This is a frequent source of confusion because platforms often apply DMCA-style notice procedures to non-copyright complaints as a matter of company policy. That is a business choice, not a legal requirement. No statute obligates a platform to remove a trademarked listing or a defamatory post on notice. The DMCA’s procedural machinery belongs exclusively to copyright, and its conditions, including the designated agent and repeat infringer policy, buy protection only against copyright claims.

Q: What must a service do from day one to qualify for the DMCA safe harbor?

A new service should treat the safe harbor as a compliance project that begins before launch. The §512(i) threshold conditions come first: designate an agent to receive notifications of claimed infringement and register that agent with the Copyright Office, and adopt and reasonably implement a policy for terminating the accounts of repeat infringers. Then the service must build the operational capacity the conditions assume: a system for receiving proper notices, a process for expeditious removal or disabling of access to identified material, and internal controls that prevent disqualifying knowledge and the combination of financial benefit and the right and ability to control infringing activity. Nothing comparable exists under section 230, which protects from the first day of operation with no prerequisites. The startup’s practical lesson is that copyright compliance has a cost that defamation compliance does not, and that cost is the price of a conditional defence.

Q: Is there a counter-notice procedure under section 230 like the DMCA’s?

No. The DMCA’s counter-notice and put-back procedure at 17 U.S.C. 512(g) is specific to copyright and exists because the DMCA’s design anticipates mistaken or abusive removals. A user whose material was taken down on a copyright notice can file a counter-notice, and the service must restore the material within a set window unless the complainant files suit. Section 230 has no parallel because it imposes no removal duties in the first place. There is nothing to counter under the immunity: a service that leaves content up needs no procedure, and a service that takes content down voluntarily is protected for that act by subsection (c)(2). The absence is structural, not an oversight. Counter-notice exists to correct errors made under a duty; where the statute creates no duty, the error-correction machinery has nothing to attach to. Reform proposals that would add removal duties to section 230 would need to answer this question themselves.

Q: How does FOSTA change the 230-versus-DMCA picture?

FOSTA, the Allow States and Victims to Fight Online Sex Trafficking Act of 2018, Pub. L. 115-164, is the one enacted narrowing of section 230. It added 47 U.S.C. 230(e)(5), excluding specified sex trafficking claims from the immunity, so that platforms can be liable under federal and qualifying state trafficking laws where they knowingly facilitate those offenses. FOSTA does not touch the DMCA, and it does not import notice-and-takedown mechanics into section 230. Its significance for this comparison is as the worked example of what reform looks like when Congress amends the immunity directly: a targeted exclusion for a defined category of claims, leaving the rest of the immunity’s automatic architecture untouched. Reformers who want broader changes point to FOSTA as precedent for carving out claims; defenders of the immunity point to it as the outer limit of what amendment can do without converting the regime into a conditional one.

Q: Do current reform proposals try to add DMCA-style notice rules to section 230?

Many proposals described in the reform debate use DMCA-style mechanics, which is why this comparison is the debate’s actual substance. The recurring architectures include requiring platforms to act on complaints within defined timeframes, conditioning immunity on maintaining complaint-response systems, and grading obligations by service size, all recognizably borrowed from the conditional safe harbor model. The proposals are typically described by their mechanics in this article rather than by their sponsors, because the mechanics are what the comparison evaluates. The accountability case for such proposals is that platforms face no legal consequence under the immunity for ignoring complaints about non-copyright harms. The institutional-design case against them is that copyright’s verifiable-claim structure does not transfer to defamation, harassment, or other context-dependent claims, and that the copyright system’s documented over-removal is the best evidence of what a notice regime produces at scale. Both positions are live, and the five axes are the tool for weighing them.

Q: Which regime gives the service more control over the outcome?

Section 230 gives the service far more control, because its protection does not depend on the service doing anything at all. A platform operating under the immunity decides its own moderation policies for business reasons, free from legal consequence whether it leaves content up, takes it down, or ignores it entirely. The DMCA safe harbor shifts control toward the complainant and the procedure: once a proper notice arrives, the service must act expeditiously to keep its defence, and the timeline and standards are set by statute rather than by business judgment. The service retains choices, such as whether to seek a counter-notice from the uploader, but the initiative belongs to the notice sender. This control asymmetry is why platforms built their trust and safety operations around voluntary policies rather than legal mandates: under the immunity they are free to moderate or not, while under the copyright regime the law dictates the workflow.

Q: Does either regime require a platform to moderate content?

Neither regime requires moderation, which surprises readers who assume platforms moderate because the law demands it. Section 230 imposes no duties of any kind; subsection (c)(2) protects voluntary moderation but does not command it, and a platform that moderates nothing keeps its immunity intact. The DMCA requires action only after a proper notice arrives, and only the expeditious removal or disabling of the identified material, not general content moderation. Outside the notice window, the statute demands policies and systems, not editorial judgment. Platforms moderate extensively for business reasons, advertiser expectations, and user retention, and the immunity was designed precisely to let them do so without fear that the act of moderating would create liability. The legal picture is therefore the reverse of the popular one: moderation is permitted and protected, but nowhere required.

Q: What happens legally when a service fails a DMCA safe harbor condition?

Losing the safe harbor removes a defence; it does not create liability. The service falls back to ordinary copyright principles, and the rights holder must still prove infringement, ownership, and the elements of the claim. In practice, though, the loss is often decisive, because the safe harbor was the service’s only realistic shield against claims arising from millions of user uploads. The Cox litigation is the leading illustration: the provider’s repeat infringer policy existed on paper but was not reasonably implemented, the harbor was forfeited, and the resulting exposure was enormous. Contrast section 230, where no conduct can forfeit the immunity short of falling into a statutory exclusion. The practical lesson for services is that DMCA compliance is a maintenance obligation, not a one-time setup. An agent designation that lapses, a policy that exists but is not followed, or a notice queue that goes unanswered can each independently cost the defence.

Q: Should a startup build one compliance checklist or two?

Two, kept strictly separate. The section 230 checklist is empty: the immunity attaches automatically, and the startup’s only task is to avoid conduct that would make it an information content provider, meaning it should not materially contribute to the creation or development of unlawful user content. The DMCA checklist is the compliance project: register the designated agent, adopt and reasonably implement the repeat infringer policy, build the notice intake and expeditious removal workflow, train staff on what makes a notice proper, and monitor for disqualifying knowledge and the financial benefit plus control combination. Merging them invites the two characteristic errors, treating the DMCA duties as if they applied to defamation complaints, or treating the immunity as if it covered copyright. The startup that holds the distinction cleanly will respond to a defamation letter with a business decision and to a copyright notice with a legal workflow, which is exactly the allocation the statutes intended.

Q: Does section 230’s intellectual property exclusion extend to trademarks and right-of-publicity claims?

Courts have read the exclusion at 47 U.S.C. 230(e)(2) broadly. The text preserves “any law pertaining to intellectual property,” and appellate decisions have applied it to copyright and trademark claims alike. A platform sued over counterfeit listings or the unauthorized use of a brand name cannot invoke section 230 and must defend on trademark principles. Right-of-publicity claims sit closer to the boundary and have produced mixed results, because some courts treat them as intellectual property within the exclusion while others treat them as state tort claims the section preempts. The practical point is that the exclusion is not limited to copyright. Any claim a court characterizes as intellectual property falls outside the immunity, which is why the DMCA’s copyright-only harbor leaves trademark and similar claims to be litigated under their own statutes. A service assessing a complaint should therefore ask first whether the claim sounds in intellectual property at all, before reaching for either regime.