Too Many Courts Are Letting States Take Wrecking Balls to the Internet (Roundup)

December 28, 2025 · by · in Content Regulation, Derivative Liability, Licensing/Contracts

‘Tis the season for Internet censorship. 🎄 More accurately, Internet censorship is now a four-season sport in state legislatures. There is not a stereotypical red state/blue state divide. Instead, the “divide” is between pro-censorship and anti-censorship legislators. You can count the number of anti-censorship legislators nationally on a single hand.

This asymmetry has unleashed a flood of misguided state Internet censorship laws, many of them in the “segregate-and-suppress” genre. This blog post covers the legal challenges to a dozen state censorship laws.

[My backlog ties directly to my 2-week trip to China in the second half of June. I couldn’t blog some rulings that came out during the trip, and subsequent rulings came faster than I could clear my backlog. Censorship is a big business nowadays].

Many of these cases are headed towards the Supreme Court. Don’t expect the censorship laws to get frosty receptions there. The Supreme Court’s 2025 Internet Law jurisprudence has leaned into censorship (TikTok v. Garland and FSC v. Paxton). As a result, I don’t believe the Supreme Court isn’t likely to save us from our own censorial legislators. Our only hope is to elect less censorial legislators, but that falsely assumes they are in the running. Due to the scarcity of anti-censorship legislators, the modern Internet is cooked.  📉

X v. Ellison, 2025 WL 3459641 (D. Minn. Dec. 2, 2025)

This is an anti-political “deepfakes” law. The court says X lacks Article III standing to challenge the law:

Attorney General Ellison asserts that X Corp.’s pleadings do not demonstrate that X Corp. itself ever acts with any particular mens rea when deepfakes are disseminated on X
.there is no showing in the pleadings that X Corp. has ever received a report of a deepfake through its own technology (whatever that is) or through its “partnerships with third parties” (whoever those are)
.the Court cannot conclude as a matter of law that X Corp. has plausibly demonstrated that it disseminates deepfakes knowingly or with reckless disregard


The pleadings do not demonstrate that X Corp. ever disseminates deepfakes with the specific intent to injure a candidate or influence the result of an election
.X Corp. is agnostic about a user’s motivation for disseminating the deepfake. And at no point does X Corp. allege or put forward any evidence that it allows deepfakes to continue circulating on X to injure a particular candidate or to influence the result of an election. Indeed, X Corp. alleges that it largely plays an uninvolved, backseat role when it comes to addressing deepfakes, instead ceding content moderation to X users who can offer Community Notes, comments, and replies to clarify potentially deceptive AI-generated media
.even if X Corp. is aware of a political deepfake disseminated on X, the Terms of Service make clear that X Corp. does not endorse the intent of the user who posted the deepfake
.

nowhere does X Corp. allege or show that it enters into the Terms of Service with X’s users for the purpose of disseminating deepfakes. Nothing in the Terms of Service reflects any purpose—either on the part of X Corp. or X’s users—to disseminate deepfakes.

Computer & Communications Industry Association v. Uthmeier, 2025 WL 3458571 (11th Cir. Nov. 25, 2025)

This is a challenge to Florida HB3, a segregate-and-suppress law that bans the availability of certain social media “addictive” features to minors. This is a rough ruling for free speech. (I discuss the district court ruling below).

The appeals court summarizes:

We agree with the district court that HB3 likely implicates the First Amendment by singling out protected expressive activities. We also agree that HB3 is likely content neutral because it is not facially content based, nor does it have a content-discriminatory purpose. We conclude, however, that HB3’s limited restrictions likely satisfy the intermediate scrutiny test for content-neutral regulations, so the Attorney General has made a strong showing he is likely to succeed on the merits.

With respect to the law’s content neutrality:

HB3 is facially content neutral. It restricts social media platforms’ ability to contract with children under a certain age if those platforms include certain addictive features. Neither its definition of “social media platform” nor of “addictive features” makes any reference to the type of content involved
.HB3 defines social media platforms by reference to a form of expression, not a subject matter.

The court talks about parental control as a less-restrictive alternative:

Florida should allow parents to oversee their children’s social media use, the district court opined, rather than stepping in and restricting access itself. The court’s lengthy discussion of alternatives already available to parents only highlights Florida’s purpose in enacting HB3: it found that parental controls were not working

And what about the law’s age authentication mandate? The court says FSC v. Paxton blessed it:

We acknowledge that the age verification requirement imposes some burden on adult speech (thus implicating First Amendment scrutiny in the first place). But the Supreme Court has recently clarified that age verification does not automatically trigger strict scrutiny because it does not constitute a “ban on speech to adults.”

When FSC v. Paxton first came out, too many experts initially took the position that the damage wasn’t too bad because the opinion purportedly cabined its effects to adult content. Opinions like this expose that fallacy. The court delinks the adult content predicate and unleashes unrestrained censorship.

Judge Rosenbaum dissents, saying HB3 is “plainly unconstitutional on its face.” As you can imagine, I think she’s right and the majority is wrong.

She writes:

the Act purports to regulate the speech of everyone who uses the covered social media websites. For minors, it acts as a categorical ban on speech (and access to speech) on covered social media platforms. And it forces the platform to demand identifying information from all users, including adults. In doing so, it chills countless users’ speech on deeply personal, political, religious, and familial matters—reaching the heart of what the First Amendment was designed to protect in the first place.

My Segregate-and-Suppress paper raises these objections and many others.

She disagrees the law is content-neutral:

The Act directly regulates expression in at least two ways. First, the Act inserts the state of Florida into users’ decisions about whether and how to speak online
.By preventing a user from accessing a platform, the State restricts the user’s access to the unique features of that platform. In doing so, the State interferes with the message, form, and content of speech the user seeks to engage in. When a user can’t join a platform, the user can’t access speech that may be available on only that platform or communicate with the audience that may be present on only that platform.

Second, the law interferes with the platforms’ editorial choices about which users can join the platform and about what content to display to those users
.Verification requirements can directly impact the content that the platform hosts.

She finds another angle to reject content neutrality:

the Act determines whether a platform is covered based on what content that platform permits. Generally speaking, if a platform involves public user-to-user speech, then the platform is covered; if it doesn’t, it’s not. That’s a content-based law, triggering strict scrutiny.

This is pernicious, especially FOR minors:

Florida prohibits minors from holding accounts on all social media platforms that allow user-to-user expression. So minors can’t participate in or contribute to discourse in the entire online forum—whether through the “form” of liking or commenting on somebody’s post, posting, contributing to a message board, or sending a direct message. That’s the effect of the law’s prohibition on minors from creating or holding accounts on these websites. Excluding individuals wholesale from a particular forum doesn’t merely regulate one “form” of speech; it regulates all forms of speech in that forum


When it comes to speaking online, the Act effectively prohibits many minors from speaking at all.

Perhaps a contributing factor to the mental and emotional distress experienced by minors today is that the government keeps telling them to STFU
.?

NetChoice, LLC v. Bonta, 2025 WL 2600007 (9th Cir. Sept. 9, 2025)

This lawsuit involves California’s “Protecting Our Kids from Social Media Addiction Act” (SB 976), a segregate-and-suppress law. The segregation requirement is inchoate because the law deputizes the AG to make rules about how to do age authentication (which it hasn’thaven’t been done), so it won’t go into effect until 2027. Meanwhile, the law contains a package of suppression requirements, some of which the district court enjoined. But the law’s most important suppression requirement is to ban personalized feeds for minors, which the Ninth Circuit doesn’t enjoin. This permits the government to control how publishers publish content and readers consume content.

The court summarizes the statute:

An addictive feed consists of media shared by other users in which that media is shown to a user based on the user’s unique information and online activity. This means that, with some exceptions, covered platforms cannot use any information provided by minor users to decide what content to show them


the more an algorithm implements human editorial directions, the more likely it is to be expressive for First Amendment purposes. An algorithm that promotes a platform’s own message to users is likely to be protected speech. [TikTok v. Garland] Such an algorithm, after all, is not unlike traditional media curated by human editors. [Moody]


On the other hand, an algorithm that “respond[s] solely to how users act online,” merely “giving them the content they appear to want,” probably is not expressive. [Moody, including Barrett’s concurrence]. Personalized algorithms might express a platform’s unique message to the world, or they might reflect users’ revealed preferences to them.

This opinion is obviously ignorant of media theory. It assumes users’ preferences are static and not shaped by the media. It assumes uers’ preferences are capable of being revealed, without distortion, by a mirror. It also ignores that it is an editorial judgment to implement an algorithm that purports only to reveal a user’s undistorted preferences.

The court disregarded the statute’s exceptions for commercial transactions and consumer reviews because “they amount to commercial solicitation.” What? How is a consumer-authored review critiquing a third-party business a “commercial solicitation” for anything? It’s literally the opposite. This is pure Calvinball.

The court continues: “California’s use of “social media” platform as statutory shorthand does not render the Act content based, since it applies to websites whether they facilitate social interaction or other forms of content.” The law singles out social media publishers for different treatment than other publishers. This is a speaker-based restriction. The court later says that not all speaker-based laws are content-based, but the court should have explained the differences more.

On an as-applied basis, the court subjects restrictions on displaying like counts to strict scrutiny. The like display restriction fails due to NetChoice v. Bonta (9th Cir. 2024).

On an as-applied basis, the court subjects the requirement that minors’ accounts be defaulted to private mode to intermediate scrutiny. “In private mode, minors cannot conform their social media habits to maximize interaction and approval of a worldwide audience. This logically serves the end of protecting minors’ mental health by reducing screentime and habit-forming platform usage
.California took a relatively nuanced approach.”

The challenge to the age authentication requirement isn’t ripe (the age authentication method won’t come online until 2027 or later).

With respect to minors’ access to algorithmic feeds, the court says:

The only speech regulated by this provision is the speech made up of the algorithms themselves. All other content—including third-party content created by other users—remains available to minors. Minors can still search for it or follow its creator, after all.

This is terrible. Among other drafting defects, the court establishes a fake dichotomy between hosting and promotion. Both are forms of publishing content, and each is an editorial expression. To say that hosting and promotion are substitutes for each other is asinine.

Thus, the court dodges the facial challenge because it believes that “all we recognize is that some personalized recommendation algorithms may be expressive, while others are not, and that inquiry is fact intensive,” so there may be constitutional applications of the law. This is grossly wrong. All algorithmic choices are expressive, and the state has stripped away some of those expressive choices from the editorial toolkit.

The court rejects the overbreadth facial challenge because the law’s potential reach is so expansive that it’s impossible to tell who is covered (the court asks: “ESPN.com? wsj.com? neopets.com? chess.com? Airbnb? Or any number of thousands of platforms”). In other words, the more poorly-drafted and overreaching the law is, the more it’s immunized from First Amendment challenges??? The Moody precedent was an unfortunate gift to bad legislators.

This opinion sucked.

Kohls v. Bonta, 2025 WL 2495613 (E.D. Cal. Aug. 29, 2025)

This case is a First Amendment challenge to AB 2839, which regulates political synthetic content a/k/a “deepfakes.” The court had previously preliminarily enjoined the law.

The court now makes the injunction permanent because “AB 2839 discriminates based on content, viewpoint, and speaker and targets constitutionally protected speech.” The court adds:

While it is true that AB 2839 has constitutional applications to the extent that defamatory or fraudulent speech falls under its umbrella, this is only because its scope is so elastic that it penalizes wholesale categories of speech, sweeping in both protected and unprotected speech. Thus, the statute’s potential unconstitutional applications would regularly outweigh its constitutional ones.

Furthermore, “AB 2839’s expansive terms capture even satire or parody videos
a mandatory disclaimer for parody or satire would kill the joke.” The court concludes: “Just as the government may not dictate the canon of comedy, California cannot pre-emptively sterilize political content.”

Thanks for passing yet another censorship law, California legislature!

Chamber of Commerce v. Lierman, No. 24-1727 (4th Cir. Aug. 15, 2025) 

This involves Maryland’s digital ad tax (“Maryland thus became the first state in the country to tax the revenues companies produce by advertising on the internet.”). The law both imposed a tax on digital ads and prohibited publishers from separately disclosing this tax “by means of a separate fee, surcharge, or line-item.” As the court says, “If companies pass on the cost of the tax, they must do so in silence—keeping customers in the dark about why prices have gone up and thereby insulating Maryland from political responsibility.”

The Fourth Circuit rejects the gag rule, saying that it would fail intermediate scrutiny (though the court thinks strict scrutiny applies). The court summarizes: “As much today as 250 years ago, criticizing the government—for taxes or anything else—is important discourse in a democratic society. The First Amendment forbids Maryland to suppress it.”

CCIA v. Uthmeier, 4:24cv438-MW/MAF (N.D. Fla. June 3, 2025)

[This decision got overturned by the 11th Circuit CCIA v. Uthmeier opinion discussed above.]

A Florida segregate-and-suppress law, H.B. 3, “prohibits some social media platforms from allowing youth in the state who are under the age of 14 to create or hold an account on their platforms, and similarly prohibits allowing youth who are 14 or 15 to create or hold an account unless a parent or guardian provides affirmative consent for them to do so.” This is one of the many anti-“addiction” segregrate-and-suppress laws.

The court summarizes:

Although this Court today finds that Florida’s challenged law is likely unconstitutional, it does not doubt that parents and legislators in the state have sincere concerns about the effects that social media use may have on youth, nor does it render parents or the State powerless to address those concerns. For example, this Order leaves in place new provisions of Florida law that require covered social media platforms to terminate any account held by a youth under 16 in the state upon the request of a parent or guardian. Instead, like other district courts around the country, this Court simply recognizes that the First Amendment places stringent requirements on the State to avoid substantially burdening speech unless the State can show that doing so is necessary to achieve its significant interests.

Of specific interest, the court says that the law triggers intermediate scrutiny. The court says the law isn’t content-based: “Social speech, or speech generated by users on social media platforms, can touch on any conceivable topic, message, or idea, and accordingly does not seem akin to the kind of category that the Supreme Court discussed in City of Austin.”

The court says the law nevertheless fails intermediate scrutiny:

Even assuming the significance of the State’s interest in limiting the exposure of youth to websites with “addictive features,” the law’s restrictions are an extraordinarily blunt instrument for furthering it. As applied to Plaintiffs’ members alone, the law likely bans all youth under 14 from holding accounts on, at a minimum, four websites that provide forums for all manner of protected speech: Facebook, Instagram, YouTube, and Snapchat. It also bans 14- and 15-year-olds from holding accounts on those four websites absent a parent’s affirmative consent, a requirement that the Supreme Court has clearly explained the First Amendment does not countenance [cite to Brown]

The court cites the Packingham decision for its rejection of categorical bans on social media:

Social media websites, according to the Court, “provide the most powerful mechanisms available to a private citizen to make his or her voice heard.” This is perhaps especially true for youth, for whom the available forums for speech are generally more limited than for adults, who tend to have freer rein to decide for themselves where to go and how to spend their time


Youth are people, not mere people-in-waiting or extensions of their parents. They have their own interests, ideas, and minds. Not only that, but they are citizens in training. The responsibilities and privileges of citizenship are significant, and our constitutional system is better served when its citizens build those muscles over time, beginning when they are young, rather than all at once the day they come of age.

The court also cites the law’s parental veto power, which the plaintiffs didn’t challenge, as sufficient to protect children’s interest. My Segregate-and-Suppress paper explains the many problems with parental consent/veto rights.

NetChoice LLC v. Carr, 1:25-cv-02422-AT (N.D. Ga. June 25, 2025)

This is a pre-FSC v. Paxton ruling, so it may not be good law any more.

The court summarizes Georgia’s segregate-and-suppress law:

The Protecting Georgia’s Children on Social Media Act of 2024 is undoubtedly aimed at protecting young people from these dangers. In attempting to do so, it also implicates at least three distinct First Amendment interests: (1) it restricts the rights of Georgia’s minors to access a vital forum of information and conversation; (2) it chills the rights of all Georgians to engage in anonymous speech online; and (3) it impedes social media platforms’ ability to communicate with their users. Those burdens cannot comport with the First Amendment.

The court emphasizes point 2: “a universal age verification requirement for Georgians using social media would potentially all but kill anonymous speech online
.requiring users to tie their online views to their identity would undoubtedly chill speech—and, likely, would disproportionately chill speech on the most controversial issues.”

Section 230 also is in play:

SB 351 requires that social media platforms prohibit “[t]he display of any advertising in the minor account holder’s account based on such minor account holder’s personal information.” But that requirement “would ultimately [result in] liability ‘for decisions relating to the monitoring, screening, and deletion of content from its network—actions quintessentially related to a publisher’s role.’. . . In such a case, section-230 immunity would likely attach.” Plaintiff is thus likely to succeed on the merits of its Section 230 challenge to the Advertising Provision, which would impose liability on social media platforms for their decisions (or lack thereof) on “whether to publish, withdraw, postpone or alter” specific advertisements to young users

The court concludes:

SB 351 purports to regulate minors’ social media use for their own good. But the State has carved out dozens of arbitrary exemptions and, in oral argument, downplayed the rigor of its parental consent requirements. The State certainly has the authority and ability to provide children and parents with education, resources, and tools to understand the detriments of social media and engage with the internet in a healthier and more productive way. But the challenged portions of SB 351 do not do so. Instead, the Act curbs the speech rights of Georgia’s youth while imposing an immense, potentially intrusive burden on all Georgians who wish to engage in the most central computerized public fora of the twenty-first century. This cannot comport with the free flow of information the First Amendment protects.

NetChoice LLC v. Fitch, 2025 WL 1709668 (S.D. Miss. June 18, 2025)

This is another pre-FSC ruling. It already reached the Supreme Court’s shadow docket, where Justice Kavanaugh wrote: “NetChoice has, in my view, demonstrated that it is likely to succeed on the merits—namely, that enforcement of the Mississippi law would likely violate its members’ First Amendment rights under this Court’s precedents
.under this Court’s case law as it currently stands, the Mississippi law is likely unconstitutional.” Despite that, even Justice Kavanaugh voted to let the law–a law he thinks is unconstitutional–go into effect. How messed up is that?

Some quotes from the (now-deprecated) opinion:

  • “The facial distinctions H.B. 1126 draws based on the message a particular digital service provider conveys, or the more subtle content-based restrictions based upon the speech’s function or purpose, render the Act content-based, and therefore subject to strict scrutiny”
  • “the Attorney General has not shown that the alternative suggested by NetChoice, the private tools currently available for parents to monitor their children online would be insufficient to secure the State’s objective of protecting children”
  • “the Act requires all users (both adults and minors) to verify their ages before creating an account to access a broad range of protected speech on a broad range of covered websites. This burdens the First Amendment rights of adults using the websites of Netchoice’s covered members”
  • “H.B. 1126 similarly requires only one parent or guardian’s consent to create an account with a covered member’s website, and it does not explain how the parent or guardian relationship is to be verified”

NetChoice v. Bonta, 2025 WL 1918742 (N.D. Cal. July 11, 2025)

The case involves a constitutional challenge to a state law parallel to the INFORM Consumers Act:

Title 1.4D requires online marketplaces to collect and periodically verify certain identification, contact, and banking information from “high-volume third-party sellers,” defined as sellers making a minimum number of sales meeting certain statutory criteria that are processed through the online marketplace. If a high-volume third-party seller fails to provide information to the online marketplace as required by Title 1.4D, the online marketplace must suspend any future sales activity of the seller pending the seller’s compliance.

The court says:

A straightforward comparison between the INFORM Act and the amended version of California Civil Code § 1749.8(b) makes clear that the two are in opposition and incompatible. Section § 1749.8(b) now requires an online marketplace to count both transactions made through the online marketplace and transactions made outside of the online marketplace when determining who is a high-volume third-party seller, while the INFORM Act requires an online marketplace to count only transactions made through the online marketplace. Both requirements cannot be satisfied simultaneously.

With respect to another provision of the law:

California Civil Code § 1749.8.9(b)(1)(D) requires online marketplaces to “[m]aintain internal written policies, systems, and staff to monitor listings in order to affirmatively prevent and detect organized retail crime.” Based on information obtained through the required monitoring, an online marketplace must take certain actions, including alerting law enforcement if the online marketplace “knows or should know that a third-party seller is selling or attempting to sell stolen goods to a California resident.” The “alerting” requirement can be fulfilled only by monitoring third-party sellers’ online content. In addition, online marketplaces must establish policies prohibiting the sale of stolen goods on the online marketplace, and impose “consequences for knowingly selling stolen goods on the online marketplace, including, but not limited to, suspension or termination of the seller’s account.” That is, the online marketplace must moderate third-party content, deciding whether to publish or to withdraw it from the platform. The Court finds that NetChoice is likely to prevail on its preemption claim grounded in Section 230 (Count 2) as to SB 1144 Section 6’s addition of California Civil Code §§ 1749.8.9(a), (b)(1)(A), and (b)(1)(D).

The other provisions of California Civil Code § 1749.8.9 do not implicate an online marketplace’s role as a publisher. For example, § 1749.8.9(b)(1)(B) requires that an online marketplace provide a mechanism to allow individuals to report suspicious conduct by third-party sellers, while § 1749.8.9(b)(1)(C) requires that an online marketplace provide a mechanism to facilitate communication with law enforcement. Those requirements do not involve publication of third-party content and thus does not trigger Section 230. While § 1749.8.9(b)(2) requires publication of the policy and mechanism mandated by other subsections, such publication is not of third-party content. Finally, § 1749.8.9(c) merely states the effective date of the new statutory section.

The State’s opposition focuses largely on provisions of SB 1144 that are not implicated by NetChoice’s Section 230 claim, and for that reason those arguments are not persuasive. The Court likewise is not persuaded by the State’s attempts to characterize controlling case law as less than definite with respect to Section 230’s bar on liability flowing from a state law requiring an online service to publish, withdraw, moderate, or monitor third-party content. In the Court’s view, Grindr, Calise, Estate of Bride, and Barnes make clear that the identified provisions of California Civil Code § 1749.8.9 fall within Section 230’s preemption clause. Finally, the State’s reliance on legislative purpose is misplaced where, as here, the text of Section 230 is clear as construed by controlling case law.

Volokh v. James, Docket No. 23-356 (2d Cir. August 1, 2025)

The constitutionality of the Hateful Conduct Law depends on how the statute is interpreted. If either substantive provision of the statute requires Plaintiffs to adopt or incorporate the State’s definition of “hateful conduct,” which includes constitutionally protected speech, then we review the statute under (at least) intermediate scrutiny, and it fails. But if Plaintiffs can comply with the Hateful Conduct Law by disclosing a content moderation policy that does not incorporate or affirmatively encompass the statute’s definition of “hateful conduct” and by providing a general mechanism for reporting content-related complaints, then we review the statute under the more relaxed standard set forth in Zauderer v. Office of Disciplinary Counsel of Supreme Court of Ohio, 471 U.S. 626 (1985), and its progeny, and the statute survives constitutional scrutiny. Whether the statute can support the latter, constitutional interpretation is a question best left to the New York Court of Appeals. We thus defer decision and CERTIFY to the New York Court of Appeals questions regarding the requirements of the statute.

National Retail Federation v. James, 2025 WL 2848212 (S.D.N.Y. Oct. 8, 2025)

New York’s algorithmic pricing disclosure law qualifies for Zauderer treatment, which it survives.

Free Speech Coalition v. Knudsen, 2025 WL 2240252 (D. Mt. Aug. 6, 2025)

Montana amended its anti-porn segregate-and-suppress law to remove any AG enforcement, so that the law can only be enforced via private rights of action (the “bounty” approach). This one “neat” trick eliminates the plaintiffs’ ability to prospectively challenge the law’s constitutionality. Nice one, Montana legislature.

District of Columbia v. Facebook, Inc., 2025 WL 2166018 (D.C. Ct. App. July 31, 2025)

This is a consumer protection action over Facebook’s Cambridge Analytica scandal:

In the District’s telling, Facebook violated the CPPA by unintentionally misleading consumers about which of their data was accessible to third-party applications through a Facebook user’s friends and about Facebook’s enforcement capabilities for auditing third-party applications. The District also alleged that Facebook made a material omission by failing to disclose to users that their data had been obtained in violation of Facebook’s policies

The appellate court reversed the trial court’s grant of summary judgment to Facebook, saying “CPPA claims based on unintentional misrepresentations need only be proved by a preponderance of the evidence, as opposed to clear and convincing evidence.”

BONUS: “Attorney General James Releases Proposed Rules for SAFE for Kids Act to Restrict Addictive Social Media Features and Protect Children Online.” I don’t have the fortitude to read this. Maybe you do.

* * *

Blog Posts on Segregate-and-Suppress Obligations