MDL 3047: the federal social media addiction case

About this page. Where the record stood as of August 28, 2026, re-seeded from the court's own docket two days after the trial ended in an entered consent judgment. The case background is as of July 25, 2026 and the trial and appellate sections as of August 18 to 20, 2026; every figure carries its date. Claims 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), documented (established by a document obtained outside a ruling, such as a settlement agreement released under public records law), scheduled (a current court date, subject to change). This page is the reader's rendering of the project's evidence-tiered claim ledger for this case; corrections land in the ledger first, then here.

The case in one paragraph

MDL 3047 gathers thousands of youth-harm lawsuits against Meta, Google, TikTok, and Snap before one California federal judge. The state attorneys general's claims survived summary judgment in June 2026, went to trial on August 18, 2026 in Oakland, and ended four days into evidence: on August 26 the court entered a consent judgment resolving the claims of 51 states and territories. Meta pays up to $16,680,647,753.21 over ten years, of which about 30 percent is contingent on Snap, TikTok and YouTube adopting the same teen time limits and reverts to Meta if they do not, plus $75 million in costs 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 resolved by it). The judgment records no admission of liability, and the personal-injury and school-district tracks continue. (As of August 28, 2026.)

What an MDL is, and is not

A multidistrict litigation is a procedural mechanism, not a class action: related federal cases are centralized before one judge for coordinated pretrial work. MDL 3047 holds three distinct tracks that coverage routinely blurs: personal-injury suits by young people and families, school-district suits, and a multistate action by state attorneys general. The bundle held 3,137 pending actions as of the August 3, 2026 Judicial Panel report, up 244 in the month since July 1, before U.S. District Judge Yvonne Gonzalez Rogers in Oakland. The formal caption is In re: Social Media Adolescent Addiction/Personal Injury Products Liability Litigation (master docket 4:22-md-03047; the consolidated state-AG case is 4:23-cv-05448, filed October 24, 2023 by a 33-state coalition).

The June 29 order, labeled

• 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: there is no factual dispute that Meta's notice and parental-consent procedures did not comply with the statute's requirements. That is a finding about procedures, not yet liability; whether COPPA even applied to the services (child-directed status, knowledge of under-13 users) remains for trial.
• 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 more time on the platform was linked to worse outcomes for teens. Disputed evidence, not a finding. It puts the functionality of the offered brakes directly inside the trial record.

The August trial, and how it ended

Jury selection began August 12, 2026 in Oakland, and opening statements and evidence began on August 18. Evidence was then heard on four days in all: August 18 (5 hours 17 minutes), August 19 (5 hours 13 minutes), August 24 (5 hours 27 minutes) and August 25 (5 hours 22 minutes), with a status conference on August 20 on trial filings, protocols and e-discovery. The states called five witnesses: Arturo Bejar, a former Meta safety engineer; Dr. Jean Twenge, the psychologist; George Volichenko and Francesco Fogu; and, on the last day of evidence, Adam Mosseri, the head of Instagram. The next morning the case was over.

The defendant was Meta alone; the other MDL defendants face the separate tracks. The trial combined the 29 participating states' consolidated COPPA claims with consumer-protection claims from four of those states (California, Colorado, Kentucky, New Jersey), before an advisory jury (Meta had withdrawn its jury demand; the judge retained the ultimate findings and was expected to decide the COPPA claims herself). The dates were the court's own: Pretrial Order No. 3 stated that "openings and evidence shall commence on Tuesday, August 18, 2026," and a July 23 case-management order set the calendar under a stacked "Bellwether Trials:" heading over the dates August 12, 2026 and February 3, 2027. Meta asked the court to stay this trial and was refused: Pretrial Order No. 6, filed July 20, 2026, says "Meta's request to stay the trial in this case is DENIED." It asked the Ninth Circuit too, and was refused there as well.

Mark Zuckerberg did not testify. He had been expected to: Meta's own counsel said so in opening statement, and the judge indicated the same. The trial ended before the states reached him. He has answered for these design decisions under oath at a trial exactly once, in the California state case in February 2026.

The settlement: what was actually entered

Court-entered, August 26, 2026. The parties filed a joint motion to enter a consent judgment, attaching a fully executed 130-page settlement agreement. The court took it up in a 49-minute session, suspended the trial, granted the motion, and signed the judgment the same day. Trial Order No. 3 records the mechanics without ceremony: "Upon filing, judgment shall be entered, the trial is deemed vacated, and the jury shall be advised that they are each discharged immediately." The judgment finds "there is no just reason for delay", enters "as a final judgment under Fed. R. Civ. P. 54 and 58", and directs that "The Clerk is ordered to enter this Judgment forthwith." Both sides waived appeal.

This is a posture worth stating carefully, because the official announcement understates it. The California Attorney General's release of the same day describes a "proposed settlement, which remains subject to court approval through entry of a consent judgment". That was accurate when it was written and was overtaken within hours by the judgment itself. A press release describes the moment it was written; an entered judgment is the record.

The money, and the part that is conditional

The agreement's own Exhibit B states the totals, and its footnote is the key to reading them: "Figures are the payment amounts per installment, not the total amounts over the full payment term." There are ten installments.

Two further sums sit outside that table: a $75 million cost fund, and a $459,293,017.80 Cambridge Analytica-era component, paid to 48 of the 51 settling states for releasing claims of that era. Read that one carefully, because the first version of this page got it wrong: the agreement names four pending Cambridge complaints as its template, but California, the District of Columbia and New Mexico are not among the releasing states, so their three pending Cambridge cases are not resolved by this settlement; only Illinois's is. The arithmetic maximum is therefore about $17.2 billion, which is where the various headline figures came from. Coverage on the day reported the total as $16.68 billion, $16.7 billion, $17 billion, $17.1 billion and $18 billion, in outlets of comparable standing.

None of them separated the guaranteed money from the contingent money, and that is the distinction that decides what Meta pays. About 30 percent of the Exhibit B maximum is a "Contingency Installment Payment", owed only once a defined trigger occurs in a given state. The trigger requires that Snap, TikTok and YouTube (the agreement's "Core Industry Members", named in terms) become bound to substantively equivalent teen time-management obligations and become subject to comparable monetary obligations to that state. If that never happens within the ten-year term, the agreement says where the money goes: those payments "shall be permanently forfeited by such Settling State and retained by Meta."

The brake, in two phases, on the same condition

The product terms are structured the same way, and this is the part of the document that matters most to this project.

Phase I applies for five years and is unconditional. Teen accounts get a default overnight block "from 12 a.m. to 6 a.m.", and a default daily limit "of 2 hours per day" counted cumulatively across the covered platforms. Note the scope: the agreement's defined term is "Meta SMP", meaning Instagram and Facebook, expressly excluding "direct messaging features linked to or included in those platforms, such as Instagram Direct or Messenger" and the rest of Meta's products, so time in messaging does not count against the limit. Phase II applies only "whenever, in the 10 years following the Effective Date, Industry-Wide Adoption has occurred" for the obligation in question, and widens the overnight block to 10 p.m. to 7 a.m. "Industry-Wide Adoption" means the same three companies being bound to equivalent obligations, by settlement, by statute, or by audited voluntary undertaking.

So the stronger brake is contingent on the competition adopting it. That is the collective-action problem this project's policy paper exists to name, priced and written into an enforceable order: Meta has conceded the brake is buildable and made its stronger setting conditional on nobody else defecting.

The order, run against this site's own test

The brake scorecard asks seven questions of any feed control. They were written for this site's readers to run against their own apps, a year before this judgment existed. Here is the court-ordered brake, held to the same seven, quoting the agreement's own defined terms.

Five of seven addressed, one of those with a real carve-out, and two gaps (immediacy, and unsupervised persistence). For a document negotiated by the party being braked, that is a remarkably close transcription of the standard, and the misses are exactly where a self-drafted brake would miss.

Two structural differences from the scorecard's own frame, and they matter more than the row-by-row. The scorecard tests a control the user holds; this order gives the teen defaults and hands the key to a parent. Nothing in it gives an adult user anything at all, and the site's argument has always been that the brake is for everyone whose "show me less" does nothing. And the order's obligations are contractual, for one company, for ten years, with the judgment stating that nothing in it establishes a standard of care anywhere else.

What it is not

An independent auditor reports to a committee of state attorneys general and must "prepare and make public an executive summary of each Final Report". But the auditor's term runs "until 120 days following the fifth Final Report", while the judgment's obligations "shall expire 10 years from the Effective Date". The audited window is roughly the first half.

And the claim that carried the whole trial, that Meta misrepresented its safety tools, gets one forward-looking sentence: "Meta is enjoined from making false, misleading, or deceptive representations regarding the effect or efficacy of safety features for Teen Users on Meta SMPs." Then the gates. It applies only to statements made after the Effective Date; it can be enforced only "with the consent of a majority of the State Committee", after 30 days notice and a chance to propose a fix; and compliance with it "shall not be subject to the independent audit". The deception count's remedy is the one term in the agreement no single state can enforce and the auditor is told not to examine.

And the judgment is explicit that it decides nothing. It records that the agreement and judgment "do not constitute an admission by Defendant of any liability, wrongdoing, or violation of any local, state, federal, or international law", and then goes further: "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."

Read that sentence twice. The most detailed set of teen design obligations any American court has entered comes with an express instruction that it is not a standard. That is the argument for writing the standard down, made by the settlement itself.

What it does not resolve

The release carves out criminal, securities, tax, antitrust and environmental liability, and two categories that matter here: "Any claims of private individuals for any types of monetary or injunctive relief", and claims by school districts and other governmental plaintiffs in JCCP 5255 or the MDL. So the personal-injury and school-district tracks of MDL 3047 continue, as do the cases against Google, TikTok and Snap, who are not parties to this judgment.

What the release reaches in rather than out is just as deliberate: the agreement enumerates fifteen state-court attorney-general cases in fourteen jurisdictions that it ends, each by caption, and obliges each settling state and Meta to move jointly within ten business days for a consent judgment in that court. The most consequential of them is Tennessee's: the Nashville trial over the "Time Spent" tools was in its fifth week of evidence when this settlement stopped it, its jury dismissed the same day. That case has its own page and ledger on this site: State of Tennessee v. Meta.

The other fourteen are in Arkansas, the District of Columbia, Iowa, Massachusetts, Mississippi, Montana, Nevada (three cases), New Hampshire, Oklahoma, Puerto Rico, Utah and Vermont. Whether the public can watch any of them close is a lottery by state. Oklahoma's is wide open: the Osage County docket for No. CJ-2023-180 is a single web request, and it already shows the settlement arriving, with a "JOINT MOTION TO VACATE ALL CASE DEADLINES" entered on August 26, the same day as the federal judgment. That is the agreement's 24-hour step, clearing the calendar; the motion that actually ends the case there comes within ten business days after it. Arkansas's portal answers a request successfully and returns nothing at all, which is worse than a refusal. Tennessee's register is free but its documents cost money. This project probed what it could on August 28 and recorded which is which in the ledger, including the five it did not probe at all and the four it reached only at the portal's front door, because "we did not look" and "you cannot look" are different facts and only one of them is about the courts.

Why 29 states and not the 33 who filed

Four states left the coalition in January 2025: Michigan, Georgia, Missouri and North Dakota. That accounts for the entire drop from the 33 who filed jointly in October 2023 to the 29 at this trial, a gap this project previously carried as an unexplained number reported by several outlets.

The reason given is cost, not merit. Michigan's Attorney General's office says the court "approved most of Meta's voluminous discovery demands upon the States and their agencies (such as MDHHS), demands with which the State of Michigan could not possibly comply, and especially so within the very short timeframe so ordered by the Court", and a filing estimated the narrowed request at roughly 700,000 documents.

That is worth sitting with, because it is a fact about litigation as an instrument rather than about this defendant. The burden of proving the case removed four state attorneys general from it before trial. For a project whose argument is that the remedy has to be written rather than litigated, the attrition rate among well-resourced public plaintiffs is evidence, not colour.

And then all four were paid. Michigan, Georgia, Missouri and North Dakota each carry a payment row in the settlement's Exhibit B, and each appears in the California Attorney General's roster of joining attorneys general. The four states that could not afford to prosecute the case are settling states in its resolution. Whether that was negotiated for them or follows from the scope of the release is not something this project has established.

Sourced to the Michigan Attorney General's office as quoted by one outlet, which is weaker than this page's usual standard and is labelled accordingly. The withdrawal filings themselves have not been read. The Ninth Circuit's caption corroborates the shape of it: Michigan and Missouri are absent and North Dakota and Georgia appear as non-appealing plaintiffs. The Exhibit B rows and the AG roster are primary.

The money that was argued for, and never awarded

This section is now history rather than forecast, and it is worth keeping, because the gap between what was demanded and what was agreed is most of what the settlement tells you.

• Requested (an estimate, not an award): roughly $1.4 trillion was Meta's own characterization, in a July 2026 filing, of its theoretical statutory exposure under the four states' penalty methodology, which counts each month a teen used the platforms more than half an hour. The 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 ever endorsed any number.
• What the states themselves asked for, which was not a number. Their own filing 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 had two components: civil penalties and disgorgement of ill-gotten gains, which they said they sought to disgorge only once. They never presented it: the trial ended before the states rested.
• A figure to stay careful with. Coverage of the opening widely reported that the states sought roughly $200 billion. That number appears in the states' own filing as Meta's 2025 revenue, cited to Meta's annual report, in a passage about the scale at which deterrence has to work. It was never a demand, and this page never printed it as one.
• What was actually agreed is in the settlement section above: up to $16,680,647,753.21 under Exhibit B, of which about 30 percent is contingent on the rest of the industry, plus $75 million in costs and the $459,293,017.80 Cambridge Analytica-era component. Against a theoretical ceiling of $1.4 trillion, the guaranteed money is roughly eight tenths of one percent. That comparison is arithmetic, not commentary: a statutory maximum is not a valuation, and the ceiling was Meta's construction of the states' method, not the states' demand.
• Documented: the first federal bellwether to approach trial, brought by a Kentucky school district, settled in May 2026 for about $27 million combined (Meta $9M, Snap $8M, ByteDance $8M, Alphabet $2.01M), per the settlement agreements obtained by Reuters through a public-records request. No admission of liability, and no enforceable design standard resulted.

The appeal Meta already lost, and what it decided

Court-held, August 10, 2026. Eight days before the trial opened, the Ninth Circuit dismissed Meta's appeal for lack of jurisdiction in a published opinion, People of the State of California v. Meta Platforms, Inc., No. 24-7032. Meta had argued that a district-court ruling refusing it Section 230 immunity is a "collateral order" it can appeal immediately, before any final judgment. The panel disagreed, and the reason is worth more than the outcome: Section 230 is "a defense to liability, not immunity from suit", so a refusal of it can be reviewed in the ordinary way after trial. TikTok's appeal and the states' conditional cross-appeals went out with it. The disposition is one word: "DISMISSED."

Two things follow that matter beyond this case. The opinion is published, so that reading of Section 230 now binds the Ninth Circuit. And it records that "Meta does not identify any constitutional interests at stake", which is why the panel treated Section 230 itself as the only candidate source of immunity. Coverage that described this appeal as a First Amendment ruling was describing something else; whether a separate appellate ruling on Meta's First Amendment arguments exists is not something this project has established.

A footnote also disposes of Meta's emergency motion to stay the trial, denied as moot. That is the second refusal to stay this trial, in the second court to consider it.

The appellate track in this case is now closed by agreement rather than by decision: the consent judgment provides that "Each Party waives all rights to appeal this Final Judgment and all orders to date arising from the Action" and gives every party five business days to withdraw what is pending. The published Section 230 holding stands; it was never the subject of the appeal that was dismissed.

What happens next

The states' case against Meta is over. Four things are not.

• The rest of MDL 3047. The personal-injury and school-district tracks are expressly carved out of the release and continue, as does the litigation against Google, TikTok and Snap, none of which is a party to this judgment. The first school-district bellwether (a Tucson case) was expected around February 2027.
• The February 2027 follow-on, whose status is genuinely unclear. A second trial for 14 more states' own state-law claims was scheduled, and a July case-management order sets the calendar under a stacked "Bellwether Trials:" heading over the dates August 12, 2026 and February 3, 2027. The settlement resolves the claims of 51 states and territories, which is more states than were ever going to be tried, so that date may be moot in whole or in part. No order this project has read says so, and a vacated trial date should not be inferred from a settlement's scope.
• Compliance. The Effective Date is the first business day after entry, so August 27, 2026; the Compliance Date is six months later. The independent auditor's term begins two months after the Effective Date, and the first public executive summary is the first real test of whether any of this happened.
• The three newer defendants, now that the count is in. A Los Angeles Unified School District case filed in March 2026 named X Corp., Discord, and Roblox alongside the original four, and the court ordered a status report. It was filed August 17: Discord is named in 23 cases, Roblox in 13, X in 6. The numbers are the least interesting part. Discord and Roblox "were not named as a party to the Master Complaint in the MDL, and discovery has proceeded without" them, they have received "none of the discovery materials in these cases", and their cases have been under stipulated stays since November 2024. X argues it does not belong here, and the judge appears to agree: the report quotes her from the bench, "X has never been a part of this case. I don't understand why it shouldn't be severed." So the defendant set is widening on complaint captions and not, so far, inside the litigation.

Why this case matters for the brake

The most consequential thing in this record is not the trillion-dollar number, and it is no longer only that whether the platform's own brakes worked as advertised became litigated evidence. It is that the case ended by writing the brake down.

A default two-hour limit. A default overnight block. A parent, not the product, holding the key to loosening it. A duty not to build the escape hatch, and a ban on nudging the user sideways when the limit bites. That is close to what this project has been arguing a brake has to be, and it now exists as an enforceable obligation on the largest of these companies.

Three things stop that being a victory lap, and all three are in the document itself. The stronger version of the brake is contingent on Snap, TikTok and YouTube adopting it, so a remedy for the collective-action problem has been made conditional on solving the collective-action problem. About 30 percent of the money runs on the same condition, and reverts to Meta if it fails. And the judgment says in terms that it does not "establish a standard of care" for anyone outside the settling states.

A settlement can bind one company for ten years. Only a written standard binds the product. That argument is in the policy paper; the test itself is at the brake scorecard, and you can file what happened on your own feed.

The full evidence ledger behind this page, every claim tiered and traced against the court record, is at the MDL 3047 distillation.

Sources

Primary, the settlement: the consent judgment as entered, Dkt 576; the fully executed settlement agreement, Dkt 572-1 (130 pages, and the source of every term above); Trial Order No. 3, Dkt 575 (trial vacated, jury discharged); the Civil Minutes for each trial day (Dkt 550, 560, 561, 571, 573, 574); and the California AG's settlement release (2026-08-26), all fetched and cached 2026-08-28. Primary, the case: the summary-judgment order, Dkt 3214 (CourtListener/RECAP, verified 2026-07-16); Meta's penalty filing, Dkt 455; the states' penalty and disgorgement reply, Dkt 473 (the filing in which $200 billion is Meta's revenue); the Ninth Circuit's published opinion of Aug 10, 2026, Dkt 541; the Civil Minutes for the opening trial day, Dkt 550; the JPML pending-actions reports (July 1 and June 1, 2026); the MDL docket and state-AG docket; the California AG's June 30 release and October 2023 filing release. Secondary: Reuters wire coverage; JURIST; Engadget; Top Class Actions.

Provenance: this page is derived from the project's per-case, evidence-tiered claim ledger for MDL 3047 (a distillation record maintained in the repository), last seeded July 16, 2026, re-seeded July 25, 2026 after a docket re-check, again August 18, 2026 from the docket documents cached that day, and again August 28, 2026 from the settlement documents, every quoted span in this page's settlement section verified verbatim against the cached consent judgment and settlement agreement. Corrections flow ledger-first: the ledger is verified against the docket, then this page is re-seeded. Figures are as of the dates shown; legal characterizations are directional and for counsel to confirm.