The social media addiction lawsuits, explained

Meta's estimate of $1.4 trillion in potential penalties is the number built for headlines. The more consequential question is much smaller: when a teenager taps "Take a Break," does anything actually stop?

Across a federal multidistrict litigation in Oakland, coordinated state-court trials in Los Angeles, enforcement cases in Santa Fe and Nashville, a regulatory proceeding in Brussels, and a courtroom in Amsterdam, that same issue keeps resurfacing. The law may never settle whether "social media addiction" is a clinical category. It can still decide whether a company accurately described its product, whether a safety control did what its name promised, and whether the system quietly worked around the user's decision. That distinction is the key to understanding the lawsuits, and to seeing the remedy they have not yet produced.

About this page. Where the record stood as of August 28, 2026, when the federal states-versus-Meta trial section was rewritten from the docket: that trial ended on August 26, 2026 in an entered consent judgment, and the settlement terms below are quoted from the executed agreement and the judgment itself, both cached. Other sections carry their own dates. New Mexico was re-seeded August 13, 2026 from the final judgment entered August 6; Brussels was re-verified August 2, 2026; Tennessee was rewritten August 28, 2026, because the same settlement ended its Nashville trial mid-evidence; the remaining U.S. court sections carry a July 25 date. Litigation here moves weekly, and every figure below carries its date. This is an explanatory and policy project, not a law firm and not a claim-matching service; if you are looking to file a claim or join a lawsuit, this is not that page. Because coverage of these cases routinely blurs what has been decided against what has merely been alleged or requested, claims below are labeled: court-held (decided as a matter of law), court-entered (a judgment the parties agreed and the court signed, which is not a finding of liability), jury-found (verdict returned, appeals possible), evidence cited (in the record, still disputed), requested (a party's proposed remedy or estimate), preliminary (a regulator's initial finding), and scheduled (a current court date, subject to change).

The map: six proceedings, not one

Press coverage tends to blur these fights into one story about a "youth safety trial." They are six distinct proceedings in three legal systems, and keeping them apart is most of the work of understanding them.

MDL 3047 is the federal hub, not the whole map. It centralizes most federal personal-injury and school-district cases (plus a multistate action by state attorneys general) for coordinated pretrial proceedings in Oakland. An MDL is not itself a class action; it is a procedural mechanism for handling related federal cases together. California's coordinated state-court cases, including the trials in Los Angeles, are a separate proceeding (JCCP 5255). New Mexico's enforcement case is separate again, in Santa Fe, and Tennessee's is separate from all of them, in Nashville, where it went to a jury in July 2026 and ended without a verdict on August 26, when the federal settlement swept it in by name. And two of the six are not American at all: the European Commission's investigation is not a lawsuit but a regulatory process under the Digital Services Act, while Bits of Freedom v. Meta is a genuine court case, brought by a Dutch foundation in Amsterdam under that same regulation. The proceedings share defendants, evidence, and themes; a verdict in one does not automatically decide the others, and a Dutch order binds nothing in an American court.

Each of the four U.S. cases has its own case file, a deeper page seeded from the project's per-case evidence ledgers, where corrections land first (each ledger is published in full, every claim tiered by the strength of its source). The two European proceedings have no case file and link straight to their ledgers instead: one because it is not a case at all, the other because a Dutch judgment is best read in its own record, where the quoted text stays in Dutch and the English beside it is marked as translation.

• MDL 3047: The big federal case: thousands of youth-harm suits against Meta, Google, TikTok, and Snap, and the states' trial against Meta ended on August 26, 2026 in an entered consent judgment.
• K.G.M. v. Meta and Google: The California state cases: about 1,600 suits, whose first test verdict hit Meta and Google for $6 million, now on appeal.
• State of New Mexico v. Meta: New Mexico won $942 million from Meta and a five-year order changing how minors' accounts work, but the judge refused to touch the algorithm.
• State of Tennessee v. Meta: Tennessee tried Meta over a safety tool the State says Meta knew was broken and kept anyway, until the August 26, 2026 settlement ended the trial in its fifth week, before any verdict.
• Bits of Freedom v. Meta (a court, but a Dutch one): A Dutch court ordered Meta to make a chosen non-profiled feed survive closing and reopening the app, which is this project's own test, imposed as law.
• The EU DSA proceedings (a regulator, not a court): The EU is scoring the brakes directly: four preliminary findings say Meta's and TikTok's time limits and parental controls do not work.

MDL 3047: the federal case against Meta, TikTok, YouTube, and Snap

The federal MDL, before U.S. District Judge Yvonne Gonzalez Rogers in the Northern District of California, held 3,137 pending actions as of the Judicial Panel's August 3, 2026 report, up 244 in the month since July 1.

The legal theory is product design, not bad content. The plaintiffs allege the companies intentionally built features (infinite scroll, autoplay, variable-reward notifications, engagement-ranked feeds) to keep young users compulsively returning, and misrepresented the safety of those products. No single post is on trial; the delivery machine is. Not every design theory survived: earlier rulings in the MDL narrowed parts of the case on Section 230 grounds, which treat some algorithmic recommendation as protected publishing. Court-held, and as of August 2026 described by a federal appeals court rather than only by coverage: the district court ruled that Section 230 bars some claims "only to the extent that they concern allegedly defective platform features targeting Meta's role as a publisher of third-party content", and on a later set of claims found Section 230 to be "a fairly significant limitation" while letting them proceed. The order numbers and dates are still not in this project's hands. The claims that keep advancing are the ones about deception and about the platform's own controls.

The most consequential ruling so far came at the end of June 2026, and its two halves deserve separate labels:

• Court-held: Meta's motion for summary judgment on the states' deception and unfair-practices claims was denied; those claims go to trial. The court also gave the states a narrower pretrial win on COPPA, the federal children's-privacy statute: Meta's notice and parental-consent procedures did not comply with the statute's requirements. That is a finding about procedures, not yet liability; the states must still establish that COPPA applied to the services at issue, including disputed questions about child-directed services and Meta's knowledge of under-13 users.
• Evidence cited: in the same order (at page 23; verified directly on the public docket), the court found Meta's own documents could support the theory that its time-restriction tools were a "public relations stunt," deployed while the company knew that more time on the platform was linked to worse outcomes for teens. That is disputed evidence, not a final finding. But mark what it means for the rest of this page: whether the platform's brakes actually worked as advertised is not a metaphor imposed on this litigation from outside. It is already inside the trial record.

Full case file, with the docket sources: MDL 3047.

The Meta trial that ended in a judgment: August 2026, Oakland

The proceeding opened on August 18, 2026 and combined two things: the 29 participating states' consolidated COPPA claims, and consumer-protection claims selected from four of those states (California, Colorado, Kentucky, New Jersey). The defendant was Meta alone; the other MDL defendants face separate personal-injury and school-district tracks. Meta asked to stay this trial and was refused twice, in two courts: the trial judge on July 20 ("Meta's request to stay the trial in this case is DENIED"), the Ninth Circuit on August 10, denying an emergency stay as moot in the same opinion that threw out its appeal.

Evidence was heard on four days, before an advisory jury. The states called Arturo Bejar, the former Meta safety engineer; Dr. Jean Twenge; George Volichenko and Francesco Fogu; and on August 25, Adam Mosseri, the head of Instagram. The next morning it was over.

Court-entered, August 26, 2026. The parties filed a joint motion to enter a consent judgment attaching a fully executed 130-page settlement. In a 49-minute session the court suspended the trial, granted the motion, and signed the judgment the same day: "Upon filing, judgment shall be entered, the trial is deemed vacated, and the jury shall be advised that they are each discharged immediately." Both sides waived appeal. It resolves the claims of 51 states, commonwealths and territories, a wider group than the 29 whose claims were being tried, and it reaches past this courtroom: the agreement lists fifteen state-court AG cases in fourteen jurisdictions that it also ends, by name, including Tennessee's Nashville trial, which was in its fifth week of evidence (its jury was dismissed the same day) and cases in Massachusetts, the District of Columbia, Utah, Vermont, New Hampshire, Oklahoma, Iowa, Mississippi, Montana, Arkansas, Nevada and Puerto Rico.

Note the posture, because the announcement understated it. The California Attorney General's release that day describes a "proposed settlement, which remains subject to court approval through entry of a consent judgment". That was true when written and overtaken within hours. What exists is an entered final judgment, not a proposal.

What Meta agreed to, and the condition attached to most of it

The money. The agreement's Exhibit B sets ten annual installments totalling a maximum of $16,680,647,753.21, plus a $75 million cost fund and a $459,293,017.80 Cambridge Analytica-era component paid to 48 of the settling states for releasing claims of that era (the pending California, D.C. and New Mexico Cambridge cases are not among them and are not resolved by this settlement). About 30 percent of the Exhibit B maximum, $5.02 billion, is contingent: Meta owes it only if Snap, TikTok and YouTube become bound to equivalent teen time-management obligations and comparable payments. If that never happens in ten years, those payments "shall be permanently forfeited by such Settling State and retained by Meta."

The brake, and this is why the settlement belongs on this site rather than only in the news. Teen accounts get a default daily limit "of 2 hours per day" across Instagram and Facebook (the agreement's covered platforms; messaging and Meta's other products are excluded) and a default overnight block "from 12 a.m. to 6 a.m.", loosened only with a parent's approval. Meta must stop teens using adjacent surfaces to "effectively circumvent" the limit, and when the limit bites its apps "shall not recommend or otherwise suggest that the Teen User utilize a different Meta SMP or engage in messaging". Like and reaction counts are off by default for teens; cosmetic-procedure filters are disabled; a non-personalized feed must be offered, re-offered every 90 days, with neither answer preselected.

That is close to what this site has argued a brake must be: default-on, hard to circumvent, and not quietly undone by the product. And the stronger version of it is conditional. The wider overnight block, 10 p.m. to 7 a.m., applies only "whenever, in the 10 years following the Effective Date, Industry-Wide Adoption has occurred", meaning the same three competitors are bound to equivalent rules. A remedy for the collective-action problem, conditioned on the collective-action problem being solved first.

What it does not do

It records no admission of liability. It carves private individuals' claims and school-district claims out of the release, so the rest of MDL 3047 continues against Meta and against Google, TikTok and Snap, who are not parties. Its independent auditor publishes executive summaries for the first five years of a ten-year judgment. The deception claims that carried both trials get one forward-looking sentence, and it is the least enforceable term in the document: Meta is enjoined from false or misleading claims about teen safety features, but only prospectively, only enforceable with the consent of a majority of a state committee after 30 days notice, and the agreement says compliance with that term "shall not be subject to the independent audit". And it says, in the judgment the court signed: "Nothing in this Consent Judgment or the Agreement shall be construed to apply to, establish a standard of care for, or serve as precedent in any non-participating U.S. state or any international jurisdiction whatsoever."

The most detailed teen-design order an American court has entered comes with an instruction that it is not a standard. That is this site's argument, made by the document.

Why 29 and not the 33 who filed. Four states left the coalition in January 2025: Michigan, Georgia, Missouri and North Dakota. Michigan's Attorney General's office attributes the departure to the cost of discovery, saying the court "approved most of Meta's voluminous discovery demands upon the States and their agencies", demands the State could not comply with in the time allowed, and a filing put the narrowed request at roughly 700,000 documents. That accounts for the whole drop from 33 to 29. It is worth noticing what it says about this litigation independently of who wins: the burden of proving the case was itself enough to remove four state attorneys general from it. All four are settling states in the August 2026 judgment, with payment rows in Exhibit B.

The trillion-dollar fight is really a counting fight

None of this was ever decided, because the trial ended before the states presented a request. It is kept because the numbers are still in circulation, and because what the settlement paid against them is the most informative comparison available.

The headline number was an estimate, and not anybody's request: roughly $1.4 trillion is Meta's own estimate, disclosed in a July 2026 filing, of its potential statutory exposure under four states' proposed penalty methodology. Meta's filing calls the demanded remedies "entirely unmoored" from the claimed practices, a sanction with "no analog in the history of consumer protection enforcement." No court has endorsed any number.

The states have not named a number, and the one widely reported as theirs is not. Their own reply of July 13, 2026 says they "will present their final requests for monetary relief at trial" and that they "do not seek double recovery". The ask has two components rather than one: civil penalties and disgorgement of ill-gotten gains, which they say they seek to disgorge only once. Coverage of the opening reported that the states seek roughly $200 billion. That figure is in the states' filing, describing Meta as a company that accrued more than $200 billion in revenue in 2025, cited to Meta's annual report, in a passage about the scale at which deterrence has to work. It is possible the states also said something like it at an August hearing, whose transcript is not publicly mirrored. Until that transcript is read, this page states what the filings show and not what the headlines say.

The more interesting issue is the counting rule underneath. Statutory penalties are multiplication problems: a maximum dollar amount per violation, times the number of violations. So the decisive question is what counts as one violation. One misleading statement? One affected user? One user-month? One day during which a control failed? The answer can turn millions into billions or trillions. New Mexico's jury, for scale, counted 75,000 violations at $5,000 each. The $1.4 trillion estimate is therefore not only a damages argument; it is a dispute over how misconduct at digital scale should be counted, and that dispute outlived this trial: the consent judgment settles a case, not a counting rule.

What it settled for. The guaranteed money in the August 2026 judgment, $11.66 billion over ten years, is roughly eight tenths of one percent of the $1.4 trillion ceiling. That is arithmetic and not commentary: a statutory maximum is not a valuation, and the ceiling was the defendant's construction of the plaintiffs' method rather than anything the plaintiffs demanded. It is still the only number that puts the two in the same frame.

The verdicts so far: $6 million and $375 million

Two juries have already ruled, in two different kinds of case.

K.G.M. v. Meta and Google (Los Angeles Superior Court) was the first bellwether from roughly 1,600 coordinated California state cases. Jury-found, March 25, 2026: a $6 million verdict, $3 million compensatory and $3 million punitive, allocated 70/30 between Meta and Google, which is why some outlets report only Meta's $4.2 million share. Mark Zuckerberg testified in person at that trial, on February 18, 2026, five weeks before the verdict. In early June (announced June 10) the trial judge denied the defendants' motions for a new trial and for judgment notwithstanding the verdict. Both defendants have now filed notices of appeal (Meta in early July; YouTube on July 13). No appellate court has reviewed the verdict, and honest reporting keeps that caveat attached. Full case file: K.G.M. and the California bellwethers.

State of New Mexico v. Meta (Santa Fe) is a different animal: a consumer-protection and child-safety case under the state Unfair Practices Act, about deceptive safety claims and child exploitation rather than addiction, and not part of the MDL. Jury-found, March 24, 2026: $375 million in civil penalties, the statutory maximum of $5,000 for each of 75,000 violations. Meta has said it will appeal. A second, judge-decided phase on public nuisance and prospective relief ended in a final judgment file-stamped August 6, 2026 (announced August 7). Court-held: the platforms are a public nuisance, Section 230 does not shield Meta from that claim, and Meta must pay a further $567 million into an abatement fund, bringing the total to $942 million. New Mexico had requested about $953 million; it did not get that figure. The court also ordered five years of supervised changes to under-18 accounts: no push notifications from 10:00 PM to 7:00 AM or during school hours, like counts hidden by default, and a 90-hour monthly usage cap across Facebook and Instagram. What it refused matters more here: it granted no relief at all against the algorithm, infinite scroll or autoplay, holding that such relief would likely violate the First Amendment and Section 230 and that industry-wide features are for the legislature. So Section 230 cut both ways in one document, and the measures that survived are the ones the court called least connected with platform content. Full case file: New Mexico v. Meta.

What the verdicts established is narrower than the headlines suggest, and still consequential: in two very different proceedings, juries were willing to accept a design-harm theory in one case and a deception theory in the other. That does not settle the national litigation. It changes the risk calculation around every trial and settlement that follows, which is visible in what happened next.

The settlements, and the trial that dissolved

The pattern the verdicts set in motion, in order. Snap and TikTok settled out of the first California bellwether around the turn of 2026, TikTok as jury selection was beginning. The first federal bellwether, brought by a Kentucky school district, settled in May 2026 for a combined figure near $27 million; the terms were not disclosed in court, but Reuters later obtained the settlement agreements through a public-records request. YouTube settled the second California bellwether confidentially (reported June 23); TikTok followed on June 30. That left Meta and Snap set for trial on July 27, 2026, and the trial never happened. Snap reached a tentative settlement announced July 20, and on July 22 the plaintiff voluntarily dismissed his remaining claims against Meta, telling the court through counsel that he had "elected to withdraw his claims" given "the overall successful result of the litigation" and the prospect of a weekslong trial.

That last event is not a settlement and should not be counted as one: it is a plaintiff walking away, and on the available reporting Meta paid nothing. Whether the dismissal was with or without prejudice is not established. Note also what the sequence costs the record: of the two California test cases tried so far, one reached a jury and the other ended with the defendants leaving the room one at a time. Bellwethers are supposed to price the risk for the thousands of cases behind them, and a case that settles out teaches the docket much less than a verdict does.

Because the Los Angeles Superior Court docket is not publicly retrievable online, the July 20 and July 22 events here rest on news coverage rather than a court filing, and are recorded at that weaker tier in the case ledger.

And then the largest one. On August 26, 2026, four days into evidence, Meta and 51 state and territorial attorneys general ended the federal trial with an entered consent judgment worth up to about $17.2 billion, described in full above. It breaks the pattern in one respect and confirms it in another: unlike every settlement before it, this one did produce a disclosed, enforceable design standard, with real defaults and an auditor. And it says in its own text that the standard binds nobody outside the settling states.

For the largest defendants, none of these numbers is financially existential. Their importance is precedential, evidentiary, and strategic. Money is moving in every direction, no liability has been admitted, and, the fact this whole site turns on: the one enforceable design standard any of it produced is expressly not sector-wide, binds one company, expires in ten years, and reserves its strongest settings for the day competitors accept the same terms. Paying and continuing is, so far, the equilibrium.

Six numbers people keep confusing

Litigation headlines routinely collapse verdicts, requests, and worst-case estimates into one category of "money sought." The six numbers are six different kinds of fact, and only one of them is a judgment. It is worth noticing that the error runs in both directions: the $1.4 trillion is the defendant's own worst case reported as the plaintiffs' demand, and the $200 billion is the defendant's revenue reported as the plaintiffs' demand. A large dollar figure in a filing attaches itself to whichever party the story is about. The settlement figure attracted a sixth variety of the same error: on the day it was entered, outlets of comparable standing reported it as $16.68 billion, $16.7 billion, $17 billion, $17.1 billion and $18 billion, and none separated the guaranteed money from the contingent money, which is the only division that determines what Meta actually pays.

So is it actually addiction?

The defense's consistent line is that "social media addiction is not an established psychiatric condition," and as a diagnostic matter that fight is real: there is no such diagnosis in the DSM, the clinical category is contested, and coverage reflects the ambiguity by calling these cases social media harm lawsuits one week and social media mental health lawsuits the next.

But that debate collapses three questions that the litigation keeps separate:

1. Is "social media addiction" a formal psychiatric diagnosis? (Contested, and possibly never resolved.)
2. Can particular designs contribute to compulsive, hard-to-control use? (An empirical question about products, not diagnostic categories.)
3. Did the company accurately describe its product's risks and the effectiveness of its controls? (A consumer-protection question courts answer all the time.)

Look at which claims actually advance. The rulings that survived summary judgment, the COPPA procedures finding, the New Mexico verdict: none of them required deciding question one. They are findings about deception and consent: the platform said a control or a safety measure did something, and the evidence indicates it did not. The DSM fight asks what to call the condition. The consumer-protection fight asks what the company said, what it knew, and whether its controls did what they promised. Courts do not need to settle the first question to answer the second.

None of this requires pretending that all social-media use is harmful, that every young user reacts identically, or that every reduction in engagement is automatically beneficial. The psychological evidence varies by person, feature, and context, and age-detection systems carry their own privacy and error risks. That is exactly why the remedy this site argues for does not dictate viewpoints, prohibited content, or a permitted number of minutes. It tests whether a user-selected control reliably produces the product state the platform said it would.

Brussels is writing the rule American courts haven't

American litigation has mostly asked who should pay after alleged harm. European regulators are asking the forward-looking question: what must the product do differently?

Preliminary: on July 10, 2026, the European Commission preliminarily found the addictive design of Facebook and Instagram in breach of the Digital Services Act (IP/26/1579). Its critique reads like a product test. On the brakes Meta already ships: Instagram's and Facebook's time-management tools, "including those activated by default for teens, can be easily dismissed and do not lead to a meaningful reduction and control of the usage of the service." On the parental controls: they "are only effective if parents and guardians possess adequate technical expertise, as well as devote effort and time to understand them effectively." The changes the Commission identified include disabling autoplay and infinite scroll by default, effective screen-time breaks, and recommender systems less driven by engagement.

Three more findings sit alongside it, and two of them are about controls rather than content. On April 29, 2026 the Commission preliminarily found Meta had failed to keep under-13s off the platforms (IP/26/920), and counted the friction: Meta's tool for reporting an under-13 user "is difficult to use and not effective, requiring up to seven clicks just to access the reporting form, which is not automatically pre-filled with the user's information." After a report is made, "there often is no proper follow-up, and the reported minor can simply continue to use the service without any type of check." On July 24, 2026 it preliminarily found TikTok's minor account settings expose minors too widely (IP/26/1679), including that even when a minor chooses a private account, "their profile photos remain accessible to anyone, including users without a TikTok account." The first of the four came on February 6, 2026, against TikTok's addictive design (IP/26/312), and it used almost the same words Brussels would later use about Meta: TikTok's time-management tools "are easy to dismiss and introduce limited friction."

All four of these findings are preliminary; Meta and TikTok can respond before any final decision or penalty, and the word "preliminarily" is doing real work in every sentence above. No fine has been imposed in any of them, and a European fine, if one ever came, would be capped at 6% of worldwide annual turnover and would require a separate non-compliance decision that has not been taken. As of August 2, 2026, this section is backed by its own evidence ledger, the EU DSA proceedings, built from the Commission's four releases and quoting them verbatim; corrections to it land there first, like every other proceeding on this page.

What matters here is the shape of the question. A regulator is scoring platform brakes by performance: too easy to dismiss, not effective, not durable. That is the same question the American deception claims keep circling, translated into prospective design language. The courtroom this page's final section says is missing is beginning to be built, just not in the United States.

Amsterdam already ordered the thing this site asks for

The closest match anywhere to this project's proposal is not American and is not famous. It is a Dutch interim judgment.

In September 2025 Stichting Bits of Freedom, a digital-rights foundation, took Meta to the Amsterdam district court under the Digital Services Act over how Facebook and Instagram treat a user's choice of a chronological, non-profiled feed. On 2 October 2025 the court largely agreed with it.

• Court-ordered: Meta Ireland must make that choice persistent, and the court spelled out what the word means: the setting has to survive the user navigating to other sections, and survive the user closing the app or website and reopening it. That is the test on this site's own homepage, written as an obligation.
• Court-ordered: the non-profiled option must be directly and easily accessible on the Instagram home page on Android, the Instagram reels section across all apps and websites, and the Facebook home page and reels section across all apps and websites.
• Court-held: resetting a user's chosen feed during ordinary use forces them to set it again and again, which causes choice fatigue and is a dark pattern prohibited by the DSA.
• The price of ignoring it: 100,000 euro per day. On 10 March 2026 the Amsterdam court of appeal upheld the judgment and raised the ceiling on those penalties from 5 million to 10 million euro, on Bits of Freedom's own cross-appeal. By then Meta had dropped part of its appeal and no longer contested that the platforms had failed to comply.

Read the limits with the result. This is interim relief, not a final merits ruling. It is Dutch law applying an EU regulation, so it binds nothing in the United States, and the First Amendment reasoning that shapes every American case on this page plays no part in it. And it is about whether a control is reachable and whether it stays set, not about ranking or about what anyone is allowed to see.

What it does establish is that the objection heard most often about a standard like this, that "controls that work" is too vague to be an enforceable obligation, is answerable. A court has already written the obligation down in operative language, and an appellate court left it standing. Whether Meta has actually complied is a separate question that nobody appears to have tested publicly, and it is one a reader can test in thirty seconds with the scorecard.

The full record, with the Dutch text of every quotation and an English translation marked as a translation, is at the Bits of Freedom v. Meta ledger.

What the lawsuits cannot fix

Every remedy in the American record shares a limit. Damages are backward-looking: they price harm after it happens, and a company can pay and keep the mechanism. Settlements are the same trade with less publicity. Even a trillion-dollar penalty would be a price, not a specification, because none of it defines what a non-harmful product must do.

On August 6, 2026 that stopped being entirely true, and the way it stopped is the argument of this whole site in miniature. New Mexico's judgment is the first American design remedy actually entered: for five years, under-18 accounts get notification blackouts overnight and during school hours, like counts hidden by default, and a 90-hour monthly cap. But the same order refused every request aimed at the feed, granting no relief on Meta's algorithms, infinite scroll or autoplay, because that relief "would likely violate the First Amendment and Section 230" and because industry-wide features "require legislative or executive branch action". So the record now contains a court that accepted the mechanism as proved, ordered the parts that do not touch content, and told the legislature to handle the rest. That is not a refutation of a design standard. It is a map of which half a court can reach, and an argument for writing the other half down. The Commission's preliminary findings remain the other partial exception, and they are still preliminary.

Three weeks later the federal settlement went further, and stopped short in a different place. The August 26, 2026 consent judgment reaches the thing New Mexico would not touch: it puts a default two-hour daily limit on teen use across Instagram and Facebook, requires that teens not be able to route around it through adjacent surfaces, forbids the app from nudging them sideways when it engages, and makes a non-personalized feed an offer Meta cannot quietly weight against. That is a specification, not a price, and it is the first one in the American record aimed at the feed itself.

Then it disclaims itself. The judgment states that nothing in it may "establish a standard of care for, or serve as precedent in any non-participating U.S. state or any international jurisdiction whatsoever". Its stronger settings wait on Snap, TikTok and YouTube accepting the same terms, and about $5 billion of its money reverts to Meta if they do not. It binds one company for ten years and audits it for five.

So the American record now contains a good description of what a brake should do, inside a document that says it is not a rule. Two courts have written versions of the same remedy, case by case, for one defendant at a time, while the thing that would apply to a product rather than a party still has to be written somewhere else.

The standard already exists in American law, for everything except feeds

The usual objection to a rule like this is that "a control that works" is too vague to enforce. That objection is weaker than it sounds. The FTC already treats a control that does not do what its label says as illegal in itself, and the general rule saying so was struck down six days before it took effect.

The Commission's theory in a run of cases is not that the company lied. It is that the control failed, and the failure was the violation:

• Alleged, in the FTC's 2023 complaint over Amazon Prime: the button labelled End Membership "did not end membership. Rather, it took the consumer to the Iliad Flow."
• Alleged, in the same complaint: people who signed up through a FireStick or Fire TV "could not cancel via these same technologies."
• Alleged, in the FTC's case against Epic Games: Epic had a working undo button, learned people were using it, then renamed it, "reduced the size of the button, moved it to the bottom of the screen, and required consumers to push and hold a button on their controller." An internal Epic email, quoted by the FTC, recorded "a roughly 35% decline in the net undo-rate."

Those are, in order, the three things a brake has to do: work, exist where the choice was made, and not be quietly degraded.

Then the general version died on a technicality. The FTC's 2024 Negative Option Rule required a cancellation mechanism as simple as enrolment. On July 8, 2025, six days before its compliance date, the Eighth Circuit vacated it in full because the Commission had skipped a required regulatory analysis. The court said it "need not address Petitioners' other substantive challenges," so it never decided whether the rule was lawful in substance, and added that it did "not endorse the use of unfair and deceptive practices in negative option marketing."

And the standard has already been drafted once, in 2011. The FTC settled with an advertising firm, Chitika, whose opt-out told users "You are currently opted out" and then quietly expired: the cookies were "set to expire after 10 days," after which they "automatically expired and disappeared from consumers' browsers." A brake that holds for a week and a half, then lets go without saying so. The consent order fixing it requires the replacement choice to "remain in effect for a minimum time period of five (5) years," to "require no more than one additional click," to show the user "the current status of their choice," to disclose "that their choice is specific to the browser they are using," and to be reachable by "a hyperlink that directly takes consumers to the mechanism" from inside every targeted ad. Duration, friction, status, reachability. Fifteen years ago, in the language of an American consent order.

One term in it is worth reading twice, because it is the compromise this argument will be offered. On scope, the order does not require the choice to carry across browsers. It requires Chitika to tell the user it does not. A regulator holding the pen chose disclosure of the gap over closure of it, which is exactly the deal a platform would rather strike about Reels, Shorts and Explore. The scorecard asks the harder question.

Two limits, because this is easy to overstate. Complaints are allegations, not findings, and the Amazon case ended in a stipulated order, which is an agreement rather than a ruling. The Chitika order binds one small company and describes cookies, not feeds. And every one of these matters is about subscriptions and cancellations. Nobody has ever pointed this reasoning at a recommendation feed. That is the gap this project is trying to close, and the useful part is that closing it does not require inventing a standard. It requires applying one that already has a docket.

The record, with the posture of each matter kept distinct, is at the FTC control-integrity ledger.

The missing piece is a standard, and this site proposes one: brake integrity. When a platform offers a control to stop, limit, reset, or redirect the feed, that control must actually work and persist: take effect promptly, cover the related surfaces, and stay set until the user changes it. A safety feature should be judged by its stopping distance, not by the presence of a button. And the standard does not require the product to overpower the user; a person can always change their mind and scroll on. It requires the product to stop fighting the user after the user has chosen to stop.

The evidence for why this standard is needed is now inside the record itself: internal documents, per the June order's account, treating the time-limit tools as a public-relations exercise while the company understood what more time on the feed meant for teenagers. Those internal documents, and the platforms' own answers under oath, are on the Congressional record too, distilled in the hearings. The full argument, including its honest limits and the current First Amendment and Section 230 terrain, is in the policy paper. The test is written out at the brake scorecard; run it on your own feed, and file what happened. The trials will decide who pays for the past. The brake is about what the product has to be next.

Sources (reviewed July 15-16 and re-checked July 25, 2026; the settlement added August 28, 2026; dispositions dated)

• MDL 3047 case count (3,137 pending / 3,312 historical, August 3, 2026): JPML pending-actions report (primary, cached 2026-08-28, along with the July and June reports this page cited earlier). The link goes to the JPML index rather than the file because the filename changes monthly; note for anyone re-fetching that the host returns a false 404 to minimal user-agent strings.
• The settlement and consent judgment (entered 2026-08-26; Exhibit B figures; Phase I/Phase II time management; the non-precedent clause): the consent judgment as entered, Dkt 576 (primary); the fully executed settlement agreement, Dkt 572-1 (primary, 130 pages, the source of every quoted term above); Trial Order No. 3, Dkt 575 (primary, trial vacated and jury discharged); California AG release (2026-08-26, official, and describing the settlement as proposed hours before it was entered). All fetched and cached 2026-08-28; every quoted span above verified verbatim against the caches.
• The trial record (four days of evidence, five witnesses, the 49-minute closing session): the court's own Civil Minutes, AG-case Dkt 550, 560, 561, 571, 573 and 574, via CourtListener/RECAP (primary, cached 2026-08-28).
• June 2026 summary-judgment ruling (order filed 2026-06-29, announced 2026-06-30; deception claims to trial; COPPA procedures finding; the "public relations stunt" language at p. 23): the order itself, Dkt 3214, via CourtListener/RECAP (primary, verified 2026-07-16); California AG press release (2026-06-30, primary); Top Class Actions (COPPA-applicability questions remain); Law.com/The Recorder (2026-06-30).
• The four states that left the coalition (Michigan, Georgia, Missouri, North Dakota, January 2025, discovery burden): Michigan Public (2026-08-12, quoting the Attorney General's office; secondary, and the only account located).
• Ninth Circuit dismissal of Meta's Section 230 appeal (published opinion, filed Aug 10, 2026; holds Section 230 is a defense to liability rather than immunity from suit, and denies Meta's emergency stay motion as moot): the opinion, via CourtListener/RECAP (primary).
• The states' own statement of their monetary ask (no total; penalties and disgorgement; $200 billion appears as Meta's 2025 revenue): the reply of July 13, 2026, Dkt 473 (primary).
• August trial structure (29-state COPPA claims plus CA/CO/KY/NJ consumer claims; Aug 12 selection, Aug 18 openings; February 2027 follow-on for 14 states): Engadget (2026-07-07); CA AG.
• Advisory jury (Meta's motion to withdraw its jury demand, Dkt 2807, heard April 15, 2026; the court's decision to empanel): joint pretrial statement, Dkt 470, via CourtListener/RECAP (primary, "the advisory jury that the Court has decided to empanel"); Law Commentary (2026-07-08). The granting ruling appears to be a minute order not available in RECAP full text.
• $1.4 trillion estimate (Meta's July 2026 filing; "entirely unmoored"; "no analog in the history of consumer protection enforcement"): Meta's penalty filing, Dkt 455, via CourtListener/RECAP (primary, verified 2026-07-16); JURIST (2026-07-07).
• K.G.M. v. Meta and Google ($6M, Mar 25, 2026; post-trial motions denied early Jun, announced Jun 10; Meta appeal early July; YouTube appeal filed Jul 13): ABC7 (2026-03-25); AP via US News (2026-07-15).
• State of New Mexico v. Meta ($375M verdict Mar 24; $567M abatement fund and injunctive order entered Aug 6, 2026, for $942M total; no algorithm relief): the final judgment (primary, No. D-101-CV-2023-02838); NM DOJ press release (primary); Santa Fe New Mexican (2026-06-16); Source NM (2026-06-17).
• First federal bellwether settlement (~$27M combined, May 2026; per-defendant terms via Reuters public-records request): The Daily Record (2026-05-18); Reuters via Yahoo Finance (2026-07).
• Second California bellwether, resolved without trial (YouTube settled Jun 23; TikTok Jun 30; Snap tentative settlement Jul 20; plaintiff dismissed Meta Jul 22): Law.com (2026-06-24); NBC News (2026-06-30); Courthouse News (2026-07-22); TechCrunch (2026-07-22). All secondary: the LA Superior Court docket is not publicly retrievable online, so no filing was obtained for the July 20 or July 22 events.
• First California bellwether settlements (Snap, then TikTok at jury selection): NPR (2026-01-27); CNN (2026-01-26).
• EU DSA preliminary findings, four, each cited to the Commission's own release (all primary; all four re-fetched, cached, and every quoted span on this page verified as an exact substring of its cache, 2026-08-02): Meta addictive design, Jul 10, 2026, IP/26/1579; Meta under-13s, Apr 29, 2026, IP/26/920; TikTok minors' account settings, Jul 24, 2026, IP/26/1679; TikTok addictive design, Feb 6, 2026, IP/26/312 (this release had never been fetched or cited by number before the 2026-08-02 pass; the reference was recovered from the Commission's own search API). The full tiered record, including what these findings do not establish, is at the EU DSA proceedings ledger.

Figures are as reported and not independently audited here; legal characterizations are directional and for counsel to confirm. Where sources conflict (the count of participating states runs from 29 in the consolidated federal case to more than 40 across all venues, depending on what is counted), this page states the specific, attributable version rather than a bare number. The summary-judgment order and Meta's penalty filing were verified directly on the public docket (CourtListener/RECAP) on July 16, 2026; other characterizations rest on the sources above.

Provenance and methodology: this explainer restates and extends section 1 of the companion policy paper (The Brake Integrity Standard) for a search-facing audience. Facts were verified against the sources above in two passes (July 15 and July 16, 2026), the second prompted by an external AI review whose factual corrections were independently re-verified before adoption; that review also contributed the four-numbers framing, the evidence-status labels, the unit-of-violation analysis, and the stopping-distance line. A same-day primary-source audit then verified the load-bearing facts against the court record itself (the docket, the summary-judgment order, Meta's penalty filing). As of July 16, 2026 the case content is sourced ledger-first: each lawsuit has an evidence-tiered claim ledger, published in full at /distillations, the per-case pages (MDL 3047, K.G.M., New Mexico, Tennessee) are seeded from those ledgers, and corrections land in the ledgers first, then flow here on dated passes. As of August 2, 2026 that is true of every proceeding on this page, including Brussels, which until then was the one section stating findings as fact with no tiered record behind them and said so; it is now backed by the EU DSA proceedings ledger, built from the Commission's four releases with every quoted span verified against a cached copy. AI-assisted throughout: the framing and judgment are the author's, the drafting was collaborative. This page is updated on dated passes, not live; the as-of line at the top is the honesty backstop.