Decided on June 23, 2026

No. 58

Cannataro, J.

[*1]

In response to increasing violence fueled by hateful rhetoric on social media, the legislature enacted General Business Law § 394-ccc (hereinafter the "Hateful Conduct Law" or "HCL") with the aim of "requiring social media networks to provide and maintain mechanisms for reporting hateful conduct on their platform[s]" (see L 2022, ch 204). Before the law became effective, the social media network plaintiffs obtained a stay of its enforcement from the federal district court, claiming that its provisions would effectively compel them to speak out against hateful conduct and otherwise chill the publication of qualifying content in violation of the First Amendment.

In reviewing that order on appeal, the United States Court of Appeals for the Second Circuit has certified three questions to us concerning the scope of the statute. The first two questions essentially ask whether a social media network can comply with the HCL without explicitly referencing its definition of hateful conduct. Applying our ordinary canons of statutory construction, we answer those questions in the affirmative. The third certified question essentially asks whether the statute requires a social media network to respond to a user report of hateful conduct. We answer that question in the negative.

I.

In 2020, a bill was introduced in the New York legislature that proposed to combat the rise of hateful rhetoric on social media by regulating the moderation policies of social media networks. The bill would have required social media networks to remove or block access to "hate speech" within 24 hours of receiving a user complaint and to "immediately" notify the person submitting the complaint about any decision, along with the reasons for the decision (see 2020 Senate Bill 7275 § 3 [b] [ii]-[iv], available at https://www.nysenate. gov/ legislation/ bills/2019/ S7275). That bill never received a floor vote.^FN1

On May 14, 2022, an 18-year-old white supremacist armed with an assault rifle opened fire on customers at a Tops supermarket in Buffalo, killing ten individuals and injuring three others. Subsequent investigation revealed that the shooter deliberately set out to kill African Americans after being radicalized by racist memes and messages on social media, and used social media both to plan his attack and to livestream it as it was occurring. Shortly before the attack began, the shooter posted a "manifesto" on the social media platform Discord in which he detailed his plan to storm the supermarket, kill the security guard, and murder "as many blacks as possible." He indicated that he had been inspired by other mass shooters whose attacks had been broadcast on social media, such as a gunman who killed 51 people at two New Zealand mosques in 2019 while live streaming his attack online. Following that model, the killer began a livestream on Twitch, an online video streaming service, as he was driving to the supermarket, and continued streaming as he began to massacre his victims. A concerned Twitch user reported the stream, and the platform terminated the broadcast about two minutes after the first victim was shot. Video and images of the attack nonetheless quickly proliferated on numerous other social media platforms. The shooter posted his manifesto and live stream with the intent to incite others to engage in racially motivated attacks.

Shortly thereafter, the legislature passed the Hateful Conduct Law with significant bipartisan support (see Bill Jacket, L 2022, ch 204 at 3-4). As its plain text makes clear, the HCL is significantly more limited in scope than the 2020 proposed legislation. The HCL contains two substantive requirements, which the parties and federal courts have referred to in this litigation as the "Report Mechanism Requirement" and the "Policy Disclosure Requirement." The Report Mechanism Requirement states:

"A social media network that conducts business in the state, shall provide and maintain a clear and easily accessible mechanism for individual users to report incidents of hateful conduct. Such mechanism shall be clearly accessible to users of such network and easily accessed from both a social media networks' application and website, and shall allow the social media network to provide a direct response to any individual reporting hateful conduct informing them of how the matter is being handled" (General Business Law § 394-ccc [2]).

The Policy Disclosure Requirement provides:

"Each social media network shall have a clear and concise policy readily available and accessible on their website and application which includes how such social media network will respond and address the reports of incidents of hateful conduct on their platform" (id. § 394-ccc [3]).

The HCL defines "hateful conduct" as "the use of a social media network to vilify, humiliate, or incite violence against a group or a class of persons on the basis of race, color, religion, ethnicity, national origin, disability, sex, sexual orientation, gender identity or gender expression" (id. § 394-ccc [1] [a])^FN2. It authorizes imposition of civil [*2]penalties up to $1,000 per day on any social media network that knowingly fails to comply with the statute, and authorizes the New York Attorney General to carry out its provisions (id. § 394-ccc [5]). The statute contains a savings clause which states that "[n]othing in this section [General Business Law § 394-ccc] shall be construed as . . . an obligation imposed on a social media network that adversely affects the rights or freedoms of any persons," including "the right of free speech pursuant to the first amendment to the United States Constitution" (id. § 394-ccc [4]). The same provision further states that the HCL shall not be construed "to add to or increase liability of a social media network for anything other than the failure to provide a mechanism for a user to report to the social media network any incidents of hateful conduct on their platform and to receive a response on such report" (id.).

Plaintiffs are Eugene Volokh, Locals Technology Inc. (Locals), and Rumble Canada, Inc. (Rumble). Volokh is a law professor and the operator of The Volokh Conspiracy, a legal blog "dedicated to maintaining a free and open marketplace of ideas." Consistent with that mission, the blog allows readers to comment on posts and participate in open forums, and "generally avoids regulating or removing visitor comments from [the] site based on viewpoint." Volokh nonetheless contends that developing and maintaining a mechanism that would allow visitors to express their concerns about hateful content directly to the blog, or transparently disclosing to such visitors the fact that it generally does not regulate such content, would be burdensome and "contrary to The Volokh Conspiracy's ethos, purpose, and mission."

Plaintiff Rumble is an online service that allows users to upload, share, and comment on video content. Rumble describes itself as "on a mission to protect a free and open internet," and considers its platform "immune to cancel culture." Rumble's terms and conditions nonetheless prohibit content that is "grossly offensive to the online community, including but not limited to, racism, anti-semitism and hatred," or content that "supports or incites violence or unlawful acts." Rumble already maintains an email address to receive user complaints, and reserves "the right to . . . remove messages which Rumble in its sole discretion determines to be undesirable, inciting violence, harmful, offensive or otherwise in violation of [its] Terms of Use." "In other words," the complaint asserts, "Rumble is the sole determinant of whether its criteria are met."

Plaintiff Locals is a wholly owned subsidiary of Rumble that allows users to upload and share content to paying or free subscribers. Subscribers can comment on content and reply to other comments. Locals describes itself as "pro-free speech" and "committed to fostering a community that is safe, respectful, and dedicated to the free exchange of ideas." The site's community guidelines prohibit "content that threatens violence against an individual or group of people" but otherwise allows users to "independently police content." Locals maintains an email address for receiving complaints but advises users that "there is no requirement that complaints will receive a response from Locals."

Shortly before the HCL was to go into effect, plaintiffs brought this action in federal court against the New York Attorney General pursuant to 42 USC § 1983 and moved to preliminarily enjoin its enforcement on a number of grounds, including that the HCL violates the First Amendment. The District Court granted the motion, concluding that plaintiffs showed a substantial likelihood of success on their First Amendment challenges (see Volokh v James, 656 F Supp 3d 431, 444, 446 [SD NY 2023]). The court reasoned that the statute requires social media networks to "speak about" and even "endorse the state's message about 'hateful conduct' " by publishing policies "expressly indicating that [their] users have a mechanism to complain about the 'hateful conduct' as defined by the" law (id. at 441). This, the court concluded, "deprives Plaintiffs of their right to communicate freely on matters of public concern without state coercion" (id. at 442 [quotation marks omitted]). The court further concluded that the statute "could have a profound chilling effect on social media users and their protected freedom of expression" because, "[e]ven though the law does not require social media networks to remove 'hateful conduct' from their websites and does not impose liability on users for engaging in 'hateful conduct,' the state's targeting and singling out of [hateful conduct] for special measures certainly could make social media users wary about the types of speech they feel free to engage in without facing consequences from the state" (id. at 445).

The Attorney General appealed to the Second Circuit, arguing that the District Court had misinterpreted the HCL. After briefing and oral argument, the Second Circuit determined that the likelihood of success of all of plaintiffs' challenges hinges "on the proper interpretation" of the statute—a question of New York law that this Court is best positioned to resolve (see 148 F4th 71, 83 [2d Cir 2025]). With respect to the Report Mechanism Requirement, the Second Circuit explained that "[i]f a social media company can comply . . . by maintaining a [*3]reporting mechanism where consumers may direct complaints—whether or not they are complaints about content meeting the State's definition of hateful conduct—and if the statute does not require social media networks to respond to individual complaints, then the Report Mechanism Requirement does not regulate speech at all" (id. at 97). "On the other hand, if social media networks are required to provide dedicated reporting mechanisms for 'hateful conduct,' as defined by the State, or if the statute requires the networks to respond to reports, then the Report Mechanism Requirement would force Plaintiffs to promote the State's view as to categories of speech that should be discouraged, reported, and responded to" (id.). "Thus understood, the Report Mechanism Requirement would compel speech, and would be subject to intermediate or strict scrutiny," which it would not satisfy (id. at 98).

The Second Circuit reasoned similarly with respect to the Policy Disclosure Requirement. If that provision "merely requires social media networks to publicly disclose their content moderation policies, whatever those policies may be, and contains no requirement that those policies actually specifically address 'hateful conduct,' as defined by the statute, then the statute requires 'purely factual and uncontroversial information about the terms under which [commercial] services will be available,' " and would pass constitutional muster (id. at 89, quoting Zauderer v Office of Disciplinary Counsel of Supreme Court of Ohio, 471 US 626, 651 [1985]). "But if the law requires disclosures that reference 'hateful conduct' or affirmatively encompass speech fitting within that definition, it would compel controversial speech," and "[w]hether evaluated under intermediate scrutiny or strict scrutiny, the Policy Disclosure Requirement would unconstitutionally burden the social media networks' First Amendment rights" (id. at 89-90).

The Second Circuit accordingly certified to this Court the following three questions:

"(1) Does a social media network comply with N.Y. Gen. Bus. Law § 394-ccc(3)'s requirement to publish a 'clear and concise policy . . . which includes how such social media network will respond and address the reports of incidents of hateful conduct on their platform' if its policy does not explicitly reference or address the content encompassed by the statute's definition of 'hateful conduct' and does not otherwise address content that encompasses this defined category?"

"(2) Does a social media network comply with N.Y. Gen. Bus. Law § 394-ccc(2)'s requirement to 'provide and maintain a . . . mechanism for individual users to report incidents of hateful conduct' if that mechanism does not explicitly reference or address the content encompassed by the statute's definition of 'hateful conduct' and does not specifically state that content meeting the statute's definition of 'hateful conduct' may be reported using that mechanism?"; and

"(3) Does N.Y. Gen. Bus. Law § 394-ccc require a social media network to provide a direct response to any individual reporting hateful conduct informing them of how the matter is being handled?" (148 F4th at 100).

II.

In interpreting statutory provisions, "our primary consideration is to discern and give effect to the Legislature's intention" (Matter of Albany Law School v New York State Off. of Mental Retardation & Dev. Disabilities, 19 NY3d 106, 120 [2012]). "As the clearest indicator of legislative intent is the statutory text, the starting point in any case of interpretation must always be the language itself, giving effect to the plain meaning thereof" (Majewski v Broadalbin-Perth Cent. School Dist., 91 NY2d 577, 583 [1998]). In conducting this review, we give words their natural meaning, while aiming to harmonize each provision with the statute as a whole and the expressed purpose of the enactment (see Lubonty v U.S. Bank N.A., 34 NY3d 250, 255 [2019]; McKinney's Cons Laws of NY, Book 1, Statutes § 98).

The Report Mechanism Requirement

The first and central requirement of the HCL is the Report Mechanism Requirement, which provides that a social media network doing business in New York "shall provide and maintain a clear and easily accessible mechanism for users to report incidents of hateful conduct" (General Business Law § 394-ccc [2]). This mandated feature must "be clearly accessible to users of such network and easily accessed from both a social media networks' application and website," and must "allow the social media network to provide a direct response to any individual reporting hateful conduct informing them of how the matter is being handled" (id. [emphasis added]).

Plaintiffs argue that this requirement is satisfied only if a social media network creates a tool exclusively for communication about hateful conduct and explains its purpose to users with explicit reference to the statute's definition of such conduct. The plain statutory text does not support that reading. Giving the legislature's words their usual and commonly understood meanings, the statute does no more than require a social media network to create and maintain a tool capable of performing two minimum basic functions: it must (1) enable users to notify a social media network of hateful conduct posted to its platform should they choose to do so, and (2) enable the network to respond to such reports should it choose to do so. Inasmuch as the HCL does not require that the tool be provided "solely" or "exclusively" for users to report incidents of hateful conduct, the statute is not offended where a social media network's reporting tool can be used, without definition or limitation, to report anything a user wishes to report.

Nor does the plain language of the HCL require that networks explain or publicize the reporting mechanism to users with reference to the statutory definition of hateful conduct. On this point, the statute states only that the mechanism itself must be "clear and easily accessible" to users; it does not by its terms require the network to encourage users to report hateful conduct or to reference or define that term for them. A simple, unobscured "report" button that provides a means for users to report any complaint, whatever its nature, constitutes "a clear and easily accessible mechanism for users to report incidents of hateful conduct."

Thus, it is only if a network's reporting mechanism cannot be used to report hateful conduct, including if the network has no reporting mechanism for any content at all, that this requirement is violated. For example, a reporting tool that is described as allowing a user to report only technical problems (e.g., broken links or server delays) would not satisfy the statute. But a tool that allows for submission of reports on a content- and viewpoint-neutral basis would comply because such a tool would allow a user to submit a report about, among other things, someone using the network "to vilify, humiliate, or incite violence against a group or a class of persons" protected by the statute (see General Business Law § 394-ccc [1] [a]). We therefore answer the Second Circuit's second certified question in the affirmative.

The Policy Disclosure Requirement

The HCL's next substantive requirement, referred to as the Policy Disclosure Requirement, provides that "[e]ach social media network shall have a clear and concise policy readily available and accessible on their website and application which includes how such social media network will respond and address the reports of incidents of hateful conduct on their platform" (General Business Law § 394-ccc [3]). The Second Circuit asks whether this requirement, like the Report Mechanism Requirement, can be satisfied without explicit reference to the statutory definition of hateful conduct. We conclude that it can.

Rather than compel networks to "specifically address hateful conduct" (148 F4th at 90), the legislature provided, more flexibly, that a network's disclosure must "include[] how such social media network will respond and address the reports of incidents of hateful conduct on their platform" (General Business Law § 394-ccc [3]). That standard can be met without explicit reference to the statutory definition of "hateful conduct," as long as the network's explanation is broad enough to encompass, or include, content falling within the HCL definition of that term (see Merriam-Webster Dictionary [11th ed. 2009], definitions of "include" and "encompass"). For example, a network might disclose that out of respect for its users' freedom of expression, it will not regulate or remove any content posted on its platform. Such a disclosure would necessarily inform users that the network will do nothing in response to reports falling within the statutory definition of hateful conduct—or any other content. Indeed, The Volokh Conspiracy blog's policy, that it "generally avoids regulating or removing visitor comments from [the] site based on viewpoint," would satisfy the HCL's disclosure requirement ^FN3. As another example, a network might [*4]disclose that it will only remove content that is illegal or threatens violence against others, and not respond to any other reports. This, disclosure, too, would sufficiently inform users "how such social media network will respond and address the reports of incidents of hateful conduct on their platform," by revealing a distinction the network intends to draw between subcategories of that statutorily-defined content. That conclusion follows from applying our ordinary rules of statutory construction.

Lending additional support for our interpretation, the Policy Disclosure Requirement is immediately followed by the savings clause, which unequivocally states that "[n]othing in this section shall be construed . . . as an obligation imposed on a social media network that adversely affects the rights or freedoms of any persons, such as exercising the right of free speech" (id. § 394-ccc [4]). The savings clause operates as a clear and express disavowal by the legislature of any intent to unconstitutionally compel speech.^FN4

Our canons of statutory construction require us to harmonize these provisions and to treat the legislature's word choice as deliberate and meaningful (see Bank of Am., N.A. v Kessler, 39 NY3d 317, 325 [2023]; Matter of DeVera v Elia, 32 NY3d 423, 436 [2018]; Matter of Burger King v State Tax Commn., 51 NY2d 614, 620-621 [1980]; People v Sharp, 107 NY 427, 455 [1887]). We therefore conclude that a network satisfies its disclosure obligation when an ordinary user can readily infer from the network's policy notice what it will do in response to a report of content that vilifies, humiliates, or incites violence against a protected class. We accordingly answer the first certified question in the affirmative.^FN5

Whether Networks Must Respond to Reports of Hateful Conduct

The final question certified by the Second Circuit asks whether the HCL "require[s] a social media network to provide a direct response to any individual reporting hateful conduct informing them of how the matter is being handled" (148 F4th at 100). The answer to that question is no. The Report Mechanism Requirement provides only that a network's reporting tool must "allow" a network to respond to a user report of hateful conduct, not that it must do so (see General Business Law § 394-ccc [2]). Similarly, the Policy Disclosure Requirement mandates that social media networks disclose "how" they will respond to and address reports of hateful conduct. It does not require them to adopt any particular policy on responding to such reports (see id. § 394-ccc [3]). No other substantive provision of the statute even arguably requires a social media network to take particular action in response to a report of hateful conduct.

III.

In their rush to condemn the law, plaintiffs and our dissenting colleagues misapprehend and overstate the legislature's objectives, which differed substantially from the 2020 proposed legislation ^FN6. The purpose of the HCR was not to regulate networks' moderation decisions or to compel networks to endorse the State's view that hateful conduct is socially destructive. It was to empower private individuals—social media users—by equipping them with the mechanisms and information they need to bring hateful conduct to networks' attention, and to identify those platforms whose policies condone such content. Specifically, the Report Mechanism Requirement ensures that social media users can express their concerns to networks about hateful conduct on their platforms, while the Policy Disclosure Requirement informs users as to what, if anything, a network will do in response to such complaints. The legislature could rationally have concluded that such mechanisms and information may be important to consumers in this age of increasingly toxic online discourse and real-world violence. Together, the requirements facilitate users' speech by giving them a channel to express complaints, and compel transparency by networks rather than endorsement, enabling users to make informed decisions when choosing among platforms based on their reporting mechanisms and procedures. A user who infers from a network's policy disclosure that it will do little or nothing to moderate hateful conduct may choose not to use, support, or associate with such a network, or otherwise use their private commercial and social leverage to renounce the network's choices. A user's understanding of a social media network's policy with respect to hateful conduct can also help them decide what actions to take in the potentially critical moments after finding a post or stream announcing an intent to perpetrate, say, a mass shooting.

Unable to identify any substantive provision nullified or rendered truly "superfluous" by our interpretation (see dissenting op at 7), plaintiffs and the dissent are left with the objection that the HCL is not forceful enough to effectively counter hateful conduct. Setting aside the oddity of that complaint coming from those who simultaneously assert that any broader interpretation would violate the First Amendment—a result the legislature explicitly sought to avoid—the policy choice of how to effectuate a particular aim is entrusted to the legislature. Our canons of construction do not counsel against modest interpretations, only absurdity and readings that would needlessly render a presumptively valid law unconstitutional (see Lubonty, 34 NY3d at 255; Overstock.com, Inc. v New York State Dept. of Taxation & Fin., 20 NY3d 586, 593 [2013]). No absurdity has been identified here, whereas the Second Circuit has indicated that plaintiffs' and the dissent's interpretation would likely render the HCL constitutionally invalid.

Accordingly, the first and second certified questions should be answered in the affirmative and the third certified question answered in the negative.

Garcia, J. (Dissenting In Part):

As one would expect from the title "Social media networks; hateful conduct prohibited," the Hateful Conduct Law (HCL) (General Business Law § 394-ccc) defines the "hateful conduct" it seeks to prohibit. After providing the definition, the statute sets out the related requirements for social media networks: they must provide a mechanism for "individual users to report incidents of hateful conduct," and publish a policy that "includes how such social media network will respond and address the reports of incidents of hateful conduct on their platform" (id. § 394-ccc [2], [3]). The majority, in response to certified questions from the Second Circuit (Volokh v James, 148 F4th 71, 100 [2d Cir 2025]), concludes that a social media network may comply with those requirements without in any way using or referring to the defined term "hateful conduct," either as part of its reporting mechanism or its mandated policy disclosure (majority op at 10-14). These answers read not as statutory analysis but as a creative defense for networks accused of failing to comply with the law. Nevertheless, despite its best efforts, the majority cannot excise the definition from the plain text. Not only will the statutory terms continue to affect networks' compliance by forcing them to either adopt that definition or accept all reports of objectionable conduct, but the uncertainty created by the majority's distortion of the statute increases the risk of arbitrary and abusive enforcement. I dissent.

Our "primary consideration" when a question of statutory interpretation is posed "is to ascertain and give effect to the intention of the Legislature" (Matter of DaimlerChrysler Corp v Spitzer, 7 NY3d 653, 660 [2006] [internal quotation marks and citations omitted]). "[O]rdinary rules of statutory construction" (majority op at 13) begin with the language itself and its accordant plain meaning (see Majewski v Broadalbin-Perth Cent. School Dist., 91 NY2d 577, 583 [1998]). The plain meaning of the text at issue here should mark the beginning and the end of our analysis.

The cornerstone of the HCL, front and center in the text, is the definition of exactly what the law aims to prohibit: " 'Hateful conduct' means the use of a social media network to vilify, humiliate, or incite violence against a group or a class of persons on the basis of" a series of enumerated protected characteristics (General Business Law § 394-ccc [1] [a]). The "hateful conduct" definition is problematic (see 148 F4th at 93 ["the statute's definition of 'hateful conduct' encompasses a swath of protected speech"]; see also X Corp. v Bonta, 116 F4th 888, 902 [9th Cir 2024] [noting that the content moderation reporting provisions of a similar California statute "compel social media companies to report whether and how they believe particular, controversial categories of content should be defined and regulated on their platforms"]). With that constitutional infirmity in mind, the Second Circuit certified two questions asking, collectively, whether a social media network can comply with the substantive provisions of the HCL by having reporting mechanisms and policy disclosures that do not "explicitly reference or address the content encompassed by the statute's definition of 'hateful conduct' " (148 F4th at 100)^FN7. The majority answers in the affirmative because "the statute is not offended where a social media network's reporting tool can be used, without definition or limitation, to report anything a user wishes to report" and "when an ordinary user can readily infer from the network's policy notice what it will do in response to a report of content that vilifies, humiliates, or incites violence against a protected class" (majority op at 10, 14). However, the plain text of the statute answers the certified questions in the negative; the desired result can only be accomplished by effectively amending the language to strip out the definition.

"[A] title or heading may help clarify or point the meaning of an imprecise or dubious provision," but "the text of the statute must take precedence over its title" (People v Page, 35 NY3d 199, 204 n 3 [2020] [internal quotation marks and citations omitted]). Here, the title and text of the statute say the same thing. "Hateful conduct," as defined in the statute, is the concern of the law's substantive provisions. The reporting mechanism requirement demands a "clear" tool for users to report "hateful conduct" (General Business Law § 394-ccc [2]). The majority insists that "it is only if a network's reporting mechanism cannot be used to report hateful conduct, including if the [*5]network has no reporting mechanism for any content at all," that the reporting mechanism requirement is violated—and that compliance may be achieved by "[a] simple, unobscured 'report' button" (majority op at 11). No explanation for how such a generic mechanism provides the necessary clarity on reporting hateful conduct is given. This "literal compliance" strategy might be a defense for a network charged with violating the terms of the statute—it is uncertain at best how a court would view the attempt to create such a loophole—but as a matter of interpretation it runs counter to both the language and purpose of the law.

Moreover, the majority's interpretation forces networks who do not want to formally adopt the definition of "hateful conduct" to accept an exponentially larger volume of reports, quite possibly exposing them to increased scrutiny by the Attorney General. That is, to enjoy the benefits of the majority's judicially crafted safe harbor, a network has only two choices in fashioning a reporting mechanism: use the definition of hateful conduct in any description of the scope of the reporting mechanism, or use an open-ended reporting mechanism for any and all conduct the user finds objectionable. For example, under the majority's view, a network could provide a mechanism for reporting the defined "hateful conduct" alone or with other specific types of conduct; or it could omit the definition and include a catchall provision, such as a mechanism for reporting "conduct that incites violence or any other user-perceived problems or objectionable conduct." However, any attempt to limit what will be accepted without using either the definition or a catchall would run afoul of even the majority's revised statute. A network that, like plaintiffs here, is singularly "dedicated to the free exchange of ideas" (see Volokh v James, 656 F Supp 3d 431, 437 [SD NY 2023]; majority op at 5-6), is forced to either use the definition or accept any and all reports. In other words, if a network wants to do the minimum required by the statute, it must use the definition.

The majority's view of the statute's policy requirement is similarly unmoored from the text, which requires a "clear and concise policy . . . which includes how such social media network will respond and address the reports of incidents of hateful conduct on their platform" (General Business Law § 394-ccc [3]). The majority's answer to the second certified question, responding that a network complies with the policy disclosure requirement "when an ordinary user can readily infer from the network's policy notice what it will do to a report of content that vilifies, humiliates, or incites violence against a protected class" (majority op at 14) uses the language of the definition of hateful conduct without acknowledging it, and assumes that it is possible for "an ordinary user" to understand an undefined concept. It also seems odd, to say the least, that a statute designed to protect the public and inform the consumer (see NY Assembly Debate on Assembly Bill A07865-A, June 1, 2022, at 183, 187-191; 148 F4th at 94) could be satisfied when a user can "infer," from a statement such as "we consider all complaints at our discretion" (see e.g. 148 F4th at 95), how a network will respond to reports of "hateful conduct."

As with the reporting mechanism requirement, the majority would require the network either to use the definition or to have a catchall policy for all reports or for all other reports not specified, as acknowledged by the Second Circuit majority (see 148 F4th at 95 [under the State's interpretation of the policy disclosure requirement, "any policy that applies to all complaints, by logical necessity, also applies to complaints of what the statute would call 'hateful conduct' (, and t)hat also leaves open the possibility that a social media network could properly describe its moderation policies using whatever categories it chose, with a catch-all provision governing anything not specifically described"] [emphasis in original]). That is, a network must have a general policy on "complaints" or one covering a subgroup of complaints plus a catchall phrase; if not it must use the definition. How this "facilitate[s]" speech (majority op at 16 [emphasis in original]) is not apparent.

The obvious goal, as the law's title makes clear, is to "prohibit" what the legislature defined as "hateful conduct" and to enforce that prohibition, the required reporting mechanism and policy disclosure must reference the legislature's definition of that term. Why else include it? As the Second Circuit acknowledged, interpreting the policy provision "to require nothing more than disclosure of any content moderation policy, without tying that policy to hateful conduct as defined by the statute, renders much of the language of the [policy disclosure requirement], and the accompanying definition of 'hateful conduct,' superfluous" (148 F4th at 95). Indeed it does, in violation of one of our most basic rules of statutory construction (see Matter of Mestecky v City of New York, 30 NY3d 239, 243 [2017] ["We have recognized that meaning and effect should be given to every word of a statute and that an interpretation that renders words or clauses superfluous should be rejected"] [internal quotation marks and citation omitted]). The plain text—and common sense—compel the conclusion that networks are required to have a mechanism that alerts the user that hateful conduct as defined by the statute may be reported and a policy that informs that user what the network will do in response to reports of such conduct.

While the text is clear and so the analysis should end there, without reference to any legislative history (see Matter of Walsh v New York State Comptroller, 34 NY3d 520, 524 [2019]), that history, in any event, supports the plain meaning. During the Assembly floor debate over the bill, its sponsor explained that the proposed enactment was aimed at concerns about the "4.75 billion posts . . . made every day on social media" and urged that networks be required to "monitor and provid[e] individuals a vehicle for recording hateful and violent intended conduct" (Assembly Debate at 183). She clarified that "[w]hat we are trying to go after here . . . with billions of posts every day, we really are trying to target this on inciting violence . . . the hateful conduct moving toward that inciting of violence" (id. at 207). Supporters of the bill noted that it "basically say[s] we want speed in getting hateful material down" (id. at 202). It is hard to understand how the majority's interpretation of the statute furthers this goal. As the Second Circuit majority noted, interpreting the statute in a manner devoid of reference to the definition of hateful conduct, and thereby permitting—and requiring, where the definition is not used—the statute's mandates to be met by generic policies and reporting mechanisms, "would do little to address the State's goal of reducing the risk of future atrocities like the Buffalo shooting" (148 F4th at 95). A policy that declares that a network "will not regulate or remove any content posted on its platform" (majority op at 12) cannot assist in removing hateful material, and a reporting mechanism that permits the reporting of any and all content on the site does not "equip[ users] with the mechanisms . . . they need to bring hateful conduct to networks' attention" (majority op at 15). The majority's answers are hard to square with the sponsor's statement that "[t]his is cutting edge . . . legislation . . . in an effort to be proactive" in response to rising "racist violence . . . increasingly targeting minority groups such as immigrants, ethnic minorities, LGBTQ" and others (Assembly Debate at 220-221).

The legislature's intent in including the definition of hateful conduct is also clear from the significant amount of the debate taken up by discussion of that definition and the concerns expressed by legislators about the term (id. at 186-188, 203-209). No mention was made in these exchanges that the defined term after which the statute was named could be eliminated from its substantive provisions by clever lawyering. Rather, the response by the Assembly sponsor of the legislation to concerns about the definition was to attempt to link the problematic "vilify" and "humiliate" terms to "incite violence" (id. at 186 ["I think it would mean vilify or humiliate as part of a hateful conduct against any of the groups or classes of persons that are listed. . . (and) it is intended with that type of conduct, as you see in that same phrase it says 'or incite violence' (s)o it's all along that same trajectory"]; see also id. at 207 ["we really are trying to target this on inciting violence"]). A good deal of the debate centered on why the definition was necessary and how its terms might be interpreted, not ignored. In sum, not only does the majority's interpretation run counter to the statute's plain meaning, but the majority also fails to "ascertain and give effect to the intention of the Legislature" (Spitzer, 7 NY3d at 660 [internal quotation marks and citation omitted]).

The majority attempts to use the statute's savings clause (majority op at 13, n 4) to preserve the statute by excising the problematic definition, summarily concluding that "the savings clause operates as a clear and express disavowal by the legislature of any intent to unconstitutionally compel speech" (majority op at 13)^FN8. This is not how a savings clause works—one does not "harmonize" a savings clause with the statute's other provisions by overriding the plain language of those substantive provisions (majority op at 13; see Tennessee Gas Pipeline Co. v Urbach, 96 NY2d 124, 134 [2001]). Any attempt to use the savings clause to preserve a statute's constitutionality by rewriting the text would "violate[] fundamental separation of powers principles" (id.).

The majority relegates any analysis of constitutional avoidance principles to a footnote, and omits consideration of, or even citation to, our seminal cases on that doctrine (see majority op at 14 n 5). It is the doctrine of constitutional avoidance, however, and not the savings clause, that underlies the Second Circuit majority's certification (see 148 F4th at 95-96, 103 [citing People v Marquan M., 24 NY3d 1 [2014]; People v Dietze, 75 NY2d 47 [1989]). After deeming the "most natural reading" of the statute to be unconstitutional (148 F4th at 90; see also [*6]656 F Supp 3d 431 [SD NY 2023]), the Second Circuit asked this Court if the State's proposed reading of the statute, that would presumably preserve its constitutionality, is possible by application of this Court's interpretative rules (148 F4th at 95-98). To answer the Second Circuit's question, those rules, including our constitutional avoidance precedent, cannot be used to cut the relevant definition from the text. While "to be sure, a statute ought normally to be saved by construing it in accord with constitutional requirements, it is basic that the very language of the statute must be fairly susceptible of such an interpretation; put otherwise, the saving construction must be one which the court may reasonably find implicit in the words used by the Legislature" (Dietze, 75 NY2d at 52 [internal quotation marks and citations omitted]; see also People v Viviani, 36 NY3d 564, 580 [2021] ["The touchstone of the avoidance canon is the text of the statute"]). The plain text of the HCL is not susceptible to the majority's interpretation.

The statute contains only two defined terms: the first describes what the statute is aimed at prohibiting, namely "hateful conduct," and the second describes who must comply, any "social media network" (General Business Law § 394-ccc [1]). The majority excises the first term from each of the statute's ensuing substantive provisions. That interpretation is not "implicit in the words used" by the legislature; it rather eliminates those words. This Court is not "at liberty to save a statute by, in effect, rewriting it in a manner that contravenes its plain wording as well as its unambiguously articulated legislative purpose" (Matter of Wood v Irving, 85 NY2d 238, 245 [1995]). For this simple reason, the constitutional avoidance doctrine is inapplicable here.

Rewriting the statute under the guise of that doctrine, as with the use of the savings clause to do the same, raises serious separation of powers concerns. "The doctrine of separation of governmental powers prevents a court from rewriting a legislative enactment . . . when the result is incompatible with the language of the statute" (Marquan M., 24 NY3d at 10). The majority's interpretation requires such significant rewriting of the statute that it is "tantamount to wholesale revision of the Legislature's enactment, rather than prudent judicial construction" (Dietze, 75 NY2d at 53). It is not a "condemn[ation]" of the statute (majority op at 15) to answer that the statute's "most natural reading" (148 F4th at 90; 656 F Supp 3d at 440-442) is the correct one—indeed the only one supported by the text—and we should not attempt to save a constitutionally suspect statute by ourselves violating constitutional norms.

Furthermore, the majority's extratextual revisions to the statute create yet another constitutional problem. The doctrine of constitutional avoidance raises "special concerns . . . in the First Amendment context" because "excessive judicial revision of an overbroad statute may lead to vagueness problems" (Marquan M., 24 NY3d at 10). That is, should the judiciary's interpretation be accepted here, "the statutory language would signify one thing but, as a matter of judicial decision, would stand for something entirely different" so that "persons of ordinary intelligence reading [the statute] could not know what it actually meant" (Dietze, 75 NY2d at 53). This is particularly dangerous with respect to "statute[s] regulating speech" because "it chills the expressive freedom of those who believe the statute means what it says and, thus, are reluctant to disobey its literal command" (id.).

Finally, the majority's revision gives free reign to the Attorney General to "enforce" its vague requirements by subpoenaing networks to monitor their compliance and by imposing significant financial penalties where those efforts are deemed insufficient. And while the majority may view its approach permitting the networks to avoid using the definition as a "modest" (majority op at 16) or at least harmless step, the Attorney General is free to enforce the statute by subpoenaing networks for all information about what reports of "hateful conduct" they have received and how they responded to each of those reports. As the Assembly Sponsor made plain, "we're not regulating at this point, we are . . . empowering, however, the Attorney General to . . . follow through and make sure the companies are complying . . . [and] enabling her to fine where they are not complying" (Assembly Debate at 207). Inherent in that power is the discretion to determine the scope of the terms "vilify" and "humiliate" (see id. at 208 [in response to a concern expressed by a legislator over "who is going to decide" what "vilify" and "humiliate" mean, the sponsor responded that it "would be . . . laid with the AG"). Consequently, if the revised statute indeed survives, it will be "too vague to obey, or to constrain arbitrary enforcement" (148 F4th at 105 [Jacobs, J., dissenting]).

It is not unprecedented for a crisis or horrific act of violence to lead to legislative overreach, however well-intentioned that reaction may be. Our role is not to rewrite such a law or to devise a strategy for avoiding compliance with its suspect terms, but to hold the constitutional line. The majority fails to do that today.

Following certification of questions by the United States Court of Appeals for the Second Circuit and acceptance of the questions by this Court pursuant to section 500.27 of this Court's Rules of Practice, and after hearing argument by counsel for the parties and consideration of the briefs and record submitted, first and second certified questions answered in the affirmative and third certified question answered in the negative. Opinion by Judge Cannataro. Chief [*7]Judge Wilson and Judges Rivera and Troutman concur. Judge Garcia dissents in part in an opinion, in which Judges Singas and Halligan concur.

Decided June 23, 2026