A different kind of speech mandate
New York's social media warning-label law is not a misinformation takedown statute. It does not tell platforms to remove anything. But it sits on the same constitutional fault line: when may a state dictate what a platform says or shows, in the name of protecting the public from information harms? That makes NetChoice's new federal lawsuit a useful test for every state that wants to regulate online speech by statute.
What happened
NetChoice, the trade group whose members include Google, Meta, Snap and TikTok, sued New York in federal court on Friday, November 13, 2026. The target is a law signed by Gov. Kathy Hochul in December 2025. It applies to platforms that use infinite scrolling, autoplay or algorithmic feeds, and it is reported to take effect in January 2027. According to NetChoice, the labels appear when users open the app and again after several hours of use. The label text is modelled on the 2023 Surgeon General advisory and says excessive use "may be associated with significant mental health harms."
NetChoice's complaint argues the law is built "to block, burden, and browbeat social media users." It says that unlike a static label on a product, "these warnings repeatedly interrupt users and block access to protected speech." The group also says the law resembles a Colorado warning-label law that a federal judge preliminarily blocked on First Amendment grounds a year ago, and that Colorado has appealed. We could not independently confirm the details of the Colorado ruling, so treat that comparison as NetChoice's framing.
The strongest case for the state
The case for New York deserves a fair statement. Governments have long required factual warnings on products that carry risk, such as tobacco and alcohol, and courts have generally tolerated them. The legislature found that heavy teen use correlates with worse mental health. The bill text paraphrases the Surgeon General's finding that adolescents who spend more than three hours a day on social media face double the risk of anxiety and depression symptoms. A label is far milder than a ban or an age gate. The Senate passed the bill 59-0 on June 12, 2025, so this is not a fringe measure.
If the question were only whether warnings are less restrictive than prohibitions, New York would have a point.
Why the doctrine still cuts against it
The problem is that the label is the state's message, delivered in the platform's own product, about contested science. The key precedent is NIFLA v. Becerra, decided June 26, 2018 by a 5-4 Supreme Court. The Court there confirmed that the Zauderer standard for compelled disclosures covers only "purely factual and uncontroversial information," and that such disclosures must not be "unjustified or unduly burdensome." It held that Zauderer had "no application" to the California notice at issue.
Two features of New York's design strain that test. First, whether "excessive use" is associated with "significant mental health harms" is an active research dispute, not the kind of settled fact found on a cigarette pack. A label that asserts a causal-sounding link on the state's behalf is closer to opinion than to disclosure. Second, the burden is not the label's existence but its repetition. The statute text we reviewed leaves the frequency and display duration to rulemaking, and it forbids platforms from obscuring the label's prominence. A warning that interrupts the feed each time it is accessed is a compelled interruption of the speech users came for.
Note that we read the introduced version of the bill, and the enacted text may differ. NetChoice also characterizes the labels as non-dismissible; the bill text we read contains no dismissal provision either way.
The connection to takedown laws
Misinformation takedown laws and warning-label laws fail for the same reason. Each hands the state a role in deciding which claims about the world are true enough to be left alone, and which need correction. With takedowns, the state pushes speech down. With labels, it pushes its own speech into the stream. Both rest on the premise that the government can reliably identify the contested-yet-settled line, and both give platforms a compliance incentive to over-apply the state's preferred framing.
There is also a practical cost to innovation. A patchwork of state labels, each with different wording, frequency and trigger conditions, forces product teams to build state-specific interfaces. Large platforms can absorb that. Smaller services and new entrants, which have no feed-algorithm team, mostly cannot. EFF's resource hub on age-gating mandates tracks how quickly these state design mandates are multiplying and notes the burden they place on users and smaller services.
A better path
Policymakers who worry about teen mental health have tools that do not require scripting platform speech:
- Fund independent, longitudinal research so any future disclosure rests on settled evidence rather than correlation.
- Support parental controls and school-based digital literacy programs, which operate without government-authored messages.
- Enforce existing consumer-protection law against concrete deceptive design practices, which regulates conduct instead of viewpoint.
- If warnings are used at all, make them a single, static, factual notice, which is the version a court is most likely to treat as ordinary disclosure.
What to watch
The immediate question is whether the court grants a preliminary injunction before the law's reported January 2027 effective date. The deeper question is whether courts treat repeated, state-authored warnings about contested harms as ordinary commercial disclosure or as compelled speech. Our view is the latter. The state has a legitimate interest in youth well-being, but a government that can require a platform to recite its message about disputed science can just as easily require it to recite a message about any other topic. That is the principle that also defeats misinformation takedown mandates, and it is worth defending when the policy goal is sympathetic.