Patterson v. Meta: Is a Social Media App a Product?

On Tuesday, September 15, 2026, the New York Court of Appeals heard argument in Patterson v Meta Platforms and three companion appeals. The plaintiffs are families of the ten people murdered at the Tops supermarket on Jefferson Avenue in Buffalo on May 14, 2022, and survivors of the attack. The defendants they are still fighting to keep in the case are Meta, Instagram, Snap, Google and YouTube, Discord, Reddit, Twitch, Amazon and 4chan. The question the state’s highest court has agreed to decide is one I get asked in my office more often every year: can a social media platform be treated as a defective product under New York law, or does federal law close that door before it opens?

I handle personal injury cases, including claims that a product was badly designed. What follows is the argument as I read it before the Court heard it, updated with what happened in the courtroom, and what the answer will mean for a New York family that believes an app hurt their child.

Update: what happened at argument (September 15, 2026)

The Court heard Patterson v Meta Platforms and the three companion appeals together on Tuesday afternoon, September 15, 2026, and kept the lawyers up for two hours and four minutes. Judge Troutman took part remotely. The plaintiffs argued as two groups with two theories: one lawyer for three of the four appeals, who said the platforms’ addictive design was a cause of the shooting independent of any content, and a second for the Patterson plaintiffs, who tied the addiction to the harmful content the algorithms fed the shooter. The New York Attorney General argued as a friend of the court on Section 230 and the First Amendment only, and said twice that the State takes no position on causation. Meta’s lawyer argued the joint issues for the internet defendants; Twitch and Amazon, and then 4chan, argued separately that they are different.

Nobody said how they would vote. The questions, though, were not evenly distributed, and they were not about the question I wrote about before argument. In two hours, no judge asked whether a social media platform is a “product.” The argument was about Section 230, the pleadings, causation and the First Amendment.

The case narrowed in the first two minutes

A judge asked the plaintiffs’ first lawyer whether she conceded that Section 230 puts any content-based claim “out of the box.” She did: “if the claim is we are suing Meta for providing white supremacist content, I agree that Section 230 bars those claims.” From that point the only claims in play were the ones that do not depend on content, and Meta’s lawyer opened by holding her to it.

The bench then spent most of her time on one demand: show us where the complaints plead that addiction, without the content, caused this shooting. “We’re not asking you to plead anything. I just want to know what’s in your complaint.” She pointed to record paragraphs on the “known propensity for violence” and on prior shootings, and she said “yes” when a judge asked whether someone addicted to cat videos would also be more likely to commit a shooting like this. Another judge told her, “I think your theory has evolved, but I don’t think that’s where it began,” and a third read the complaints back to her: “you’re feeding me white supremacist material which is altering my brain and so I commit a white supremacist shooting, at bottom, as I read your complaints, that’s what this is.” Three times she offered to re-plead.

Judge Garcia asked for the limiting principle. If addictive design is the wrong, is every harm that follows from the addiction chargeable to the company, including “I can’t function, I can’t work anymore because of my addiction to social media”? She said that is “a viable claim” if the plaintiff can prove the design caused the addiction and the addiction caused the harm. The Patterson plaintiffs’ lawyer, asked the same question later (“what would be a bridge too far under your theory?”), could offer only foreseeability and trivial harms.

Meta was pressed on the breadth of its own rule

The other side of the bench’s attention went to how far Section 230 reaches. A judge asked Meta’s lawyer what conduct of a platform is not covered by the statute; the answer was the platform’s own speech, citing this Court’s decision in Shiamili v Real Estate Group of New York, 17 NY3d 281 (2011), and the statute’s written exceptions. Judge Halligan asked whether Judge Katzmann was right, in Force v Facebook, that Section 230 has been read far beyond what Congress could have anticipated; the answer was that “the world has changed, but the statute hasn’t,” and that there is “an unbroken line of cases” the other way.

Then the hypotheticals. One judge asked: if I build an app and tell an AI to design it to do the most harm possible to the mental health of teenagers using only content it finds online, am I liable? Meta’s lawyer said he would “have to know a little bit more about it.” “What more would you need to know?” The question was not answered. Judge Cannataro asked why an “irresistible infinite feed” that keeps a user from leaving the house is a decision about how to publish rather than “an insidious design feature.” Another judge asked for a single example of a decision about how to publish that does not turn on the nature of the content, and did not get one. Meta’s position, stated on the record, is that even a complaint with no reference to content, only an addictive feed, is barred by Section 230; in the same breath its lawyer asked the Court not to reach that question, because, in his reading, these complaints use the word “content” nearly 400 times and never leave the barred category.

Causation, in the Court’s own vocabulary

The judges supplied their own cases. Is the shooting a “normal or foreseeable consequence of the situation created by the defendant’s negligence,” the test from Hain v Jamison, 28 NY3d 524 (2016), and Scurry v New York City Housing Authority, 39 NY3d 443 (2023)? Both sides claimed Kush v City of Buffalo, 59 NY2d 26 (1983): the plaintiffs for the long chain that went to a jury there, Meta for the rule that an intervening criminal act generally severs liability. When Meta’s lawyer said mass shootings are an “infinitesimal number” of the billions of interactions on these services, a judge answered with a defective Volkswagen that kills a tenth of a percent of its drivers: “doesn’t mean it’s not defective.” A judge also reminded him that surviving a motion to dismiss is not winning: it “is a long way to go after that. Maybe you’ll settle, maybe they’ll find a smoking gun in discovery.” His answer was that this is not how the Court has understood the limits of proximate cause.

The First Amendment, and “heads you win, tails you win”

Asked whether its First Amendment argument is “solely about the content,” Meta’s lawyer said he “would quibble with solely,” and then, pressed for a yes or no on whether the point depends on the Court accepting his reading of the pleadings, said, “I think that’s probably right.” Judge Halligan asked whether Meta claims First Amendment protection as a distributor of third-party speech while also claiming Section 230 because the speech is not its own. It does, and it was “happy to win either way.” A judge put the Fourth Department’s phrase to him directly: “Heads you win, tails you win, right? That’s the conundrum.” “In a case like this, that is absolutely correct, your honor.” The bench also asked whether the expressive-activity question can be decided on a motion to dismiss at all (“we don’t know how the algorithms operate here. Do we have the information to decide it?”), and, on the failure-to-warn claim, whether a warning about addictive design would be compelled speech any more than the warning on a cigarette package; Meta’s lawyer called that a hard question and asked to stay on Section 230.

The smaller defendants

Chief Judge Wilson observed that Discord, 4chan and Twitch are not pleaded as having algorithms and “strike me as really just sort of bulletin boards.” The Patterson plaintiffs’ theory against Twitch, that the shooter’s stream became Twitch’s property under its terms of service, drew the question “is that the statutory language?” and the answer “No, it’s not.” On rebuttal the plaintiffs’ lawyer conceded that “4chan is a harder case” with fewer allegations, and said she would not “fight you too hard” against a decision that reversed the legal rule and sent the case back for a defendant-by-defendant look at the pleadings, or gave leave to amend. A judge answered that the Court can read the complaints itself: “if it’s not there, what are we sending it back for?”

What I took from it

The significance of the argument, as I read it, is as follows. Three things, and none of them is a prediction.

First, this was argued as a pleading case. The content-neutral addiction theory lives in the plaintiffs’ reply briefs and in their answers from the podium more than in the four complaints, and the bench said so from several seats. In my judgment the lesson for every pending New York case against a platform is the one I would give a client tomorrow: plead the features, not the feed. Autoplay, the endless feed, the notifications at 2 a.m., the account that cannot be deleted, and the harm to the user of the account, with the company’s own research on what those features do. A complaint that reads like the story of what someone watched is a Section 230 complaint no matter what the caption says.

Second, Meta’s exposure at argument was the breadth of its own rule, not the facts of this case. It told the Court that a complaint about nothing but an addictive feed is barred, and it could not name a decision about “how to publish” that does not depend on what is published. If the Court writes on that question, that is where the writing will be.

Third, the Court has more than one narrow road, and it drew the map itself: affirm on the pleadings and causation without a broad Section 230 holding, which Meta asked for; reverse on the legal rule and decide the smaller defendants’ complaints itself, which the plaintiffs said they would not fight; or treat Section 230 immunity and tort causation as the separate questions more than one judge framed them as. The “product” question I wrote about before argument may not be reached at all.

A decision is usually a few months out. I will post when it comes down.

Quotations are from the Court’s webcast of the September 15, 2026 argument, transcribed by the firm from the webcast audio and checked line by line; the official transcript will be posted on the Court of Appeals website.

What the families allege

The complaints do not claim that a post was defamatory or that a video was illegal. They claim that the platforms were engineered to be addictive, that the recommendation algorithms fed an eighteen-year-old an escalating diet of racist and violent material because it kept him engaged, and that the companies knew the design did this and shipped it anyway. The legal theories are the familiar ones from any defective product case: design defect, failure to warn, negligence. The alternative designs the plaintiffs describe are also familiar to anyone who has argued a design case: no autoplay, a feed with an end, real parental controls, session limits, easy account deletion.

Justice Paula Feroleto in Erie County denied the platforms’ motions to dismiss in 2024. She treated the allegations as design claims, not content claims, and said the Section 230 arguments “may ultimately prove true” but could not be resolved on the pleadings.

What the Fourth Department held

On July 25, 2025, a divided panel reversed as to the social media defendants. The majority’s reasoning had three layers, and the Court of Appeals has agreed to review all three.

Section 230

The federal Communications Decency Act says a website is not treated as the publisher of content posted by its users. The majority read the complaints as depending entirely on that content: the shooter was not radicalized by an infinite scroll of cooking videos, so the theory of harm is “inextricably intertwined” with the platforms’ role as publishers. The majority said flatly that there is no strict products liability exception to Section 230, and rejected the Third Circuit’s view in Anderson v TikTok that an algorithm’s recommendations are the platform’s own speech rather than third-party content.

The First Amendment

The majority added a second lock on the same door. Citing the Supreme Court’s 2024 decision in Moody v NetChoice, it held that choosing and ranking content is editorial judgment protected by the First Amendment, even when an algorithm does the choosing. The opinion put it as a “heads I win, tails you lose” proposition for the platforms: either the algorithm is publishing third-party content and Section 230 applies, or it is the platform’s own expression and the First Amendment applies. In the majority’s view they are protected by both and “under no circumstances” by neither.

Products and causation

The majority also concluded that the platforms are not products for purposes of strict products liability, and said that even a pure addiction theory would fail on causation because a mass murder is an intervening criminal act that breaks the chain.

The two dissenting justices saw a different case. They read the complaints as targeting conduct, not content: features like autoplay, an endless feed, push notifications and the absence of parental controls are design choices that would be defective whether the content was hate speech or cat videos. In their view those features are the platform’s own creation, not third-party speech, so Section 230 does not apply, and a company that manufactures and distributes a uniform product to millions of users can be held to the same design-defect standard as any other manufacturer. Because two justices dissented on a question of law, the families had a right to appeal to the Court of Appeals without asking permission, and that is how the case reached Albany.

Why the “product” question matters so much in New York

New York’s products liability law is court-made and old. A manufacturer is strictly liable for a design defect when the product, as designed, was not reasonably safe and a safer alternative design was feasible. The test comes from Voss v Black & Decker and it has been applied to cars, saws, ladders and drugs for fifty years. It has never been squarely applied by the Court of Appeals to software, let alone to a free app whose “user” is also its raw material.

If the Court holds that a platform can be a product, three things follow for a New York plaintiff. The design-defect framework applies, with its focus on feasible alternatives, which is exactly what the plaintiffs pleaded. The claim can be tried in a New York Supreme Court rather than folded into the federal multidistrict litigation in California. And discovery into how the features were designed and tested, the internal research that has surfaced in other cases, becomes available under New York’s broad disclosure rules.

If the Court holds that a platform is not a product, the design theory does not disappear, but it becomes an ordinary negligence claim, and ordinary negligence against a platform runs straight back into Section 230 and the First Amendment analysis the majority adopted.

What each side has to win

The families’ cleanest path is the dissent’s framing: keep the case about features, not feeds. Their difficulty is the Buffalo record itself. The Fourth Department majority was right that these complaints describe the specific racist content the shooter consumed, because that is the true story of what happened. A design-only complaint about a child who lost sleep and developed an eating disorder is easier to keep clear of Section 230 than a complaint about a man who was radicalized by what he read. Whether the Court will separate the theory from the facts of this particular tragedy was where the bench spent most of the argument; the update above covers it.

The platforms’ cleanest path is Moody. If curation by algorithm is protected expression, the product label does not matter, because a state cannot impose tort liability for protected speech. Their difficulty is that the Supreme Court in Moody was talking about government laws that force platforms to carry speech, and the dissent below drew that line sharply: a private tort claim about an addictive design is “a far cry” from a state statute about content moderation.

The causation questions came, and they came in the Court’s own vocabulary; see the update above.

What this means for a New York family today

Cases against the platforms brought by school districts, parents and young adults are already pending in New York courts and in the federal multidistrict litigation, where a federal judge in California allowed several design-based claims to proceed in 2023 while dismissing content-based ones. Our parents’ guide to social media harm lawsuits covers who can sue and what the claims look like, and our note on the first bellwether trial in the federal MDL covers the federal side. A ruling for the families in Patterson would make the state court route viable for New Yorkers. A ruling against them would channel New York claims into the federal case.

If you believe an app injured your child, the answer in Patterson does not change three things you should do now. Preserve the account data, the screen-time records and the device; that evidence disappears with a factory reset. Get the diagnosis documented by a treating clinician, because the medical causation question is where these cases are actually won or lost. And do not wait for Albany: the statute of limitations on a personal injury claim in New York is three years, and for a minor it runs from the eighteenth birthday, but the evidence does not wait for anyone. Our product liability practice handles design-defect claims, and a consultation is free.

Timeline

Sources

• Patterson v Meta Platforms, Inc., 2025 NY Slip Op 04446 (4th Dept July 25, 2025), incorporating the memorandum in Salter v Meta Platforms, Inc.
• Court of Appeals, Case Issue Statements, September 2026 (Patterson v Meta Platforms and related appeals, Nos. 68 to 71)
• Eric Goldman, “Social Media Services Aren’t Liable for Buffalo Mass-Shooting,” Technology & Marketing Law Blog (July 28, 2025), quoting the majority and dissent
• Moody v NetChoice, LLC, 603 U.S. 707 (2024)
• Court of Appeals, webcast of the September 15, 2026 argument (Nos. 68 to 71); the Court’s own YouTube recording (2 hours 4 minutes), transcribed by the firm
• Hain v Jamison, 28 NY3d 524 (2016); Scurry v New York City Hous. Auth., 39 NY3d 443 (2023); Kush v City of Buffalo, 59 NY2d 26 (1983); Shiamili v Real Estate Group of N.Y., Inc., 17 NY3d 281 (2011); Force v Facebook, Inc., 934 F3d 53 (2d Cir 2019), as cited from the bench and the podium

---

Whatever the Court decides, the families in Buffalo have already forced the question that every state court will eventually have to answer. Whether the thing in a teenager’s hand at 2 a.m. is speech or a machine is not an abstraction to the people who have to live with what it did.

Call 516-750-0595 or request a free case review if you believe a social media platform’s design injured you or your child.