A federal judge in Oakland entered a consent judgment on Wednesday that forces Meta to cut off teenage users of Instagram and Facebook once they have spent two hours a day on the apps, to lock them out overnight, and to hide like counts by default. The agreement becomes effective today, starting a six-month clock on the product changes and a ten-year clock on payments that the underlying exhibits price at a maximum of $17.2 billion.

What the court entered

Chief United States District Judge Yvonne Gonzalez Rogers signed the Meta and State Attorneys General Consent Judgment on August 26, 2026, in the Northern District of California, according to the entered order filed as Document 3451 in MDL No. 3047. The order terminates Docket No. 572, the joint motion the parties filed earlier the same day, and directs the clerk to enter judgment forthwith. Magistrate Judge Peter H. Kang is also named on the caption.

The case is People of the State of California, et al. v. Meta Platforms, Inc., No. 4:23-cv-05448-YGR, folded into the multidistrict litigation captioned In re: Social Media Adolescent Addiction/Personal Injury Products Liability Litigation, No. 4:22-md-03047-YGR. The states filed their complaint on October 24, 2023. According to the joint motion, the parties negotiated for more than two years and had already begun trial when they reached terms.

Timing matters here more than usual. The settlement agreement defines the Effective Date as the first business day after the MDL court enters the consent judgment. Entry fell on Wednesday. The Effective Date is therefore today, and every deadline in the document counts forward from this morning.

Meta signed on August 25, 2026. Two signatures appear on the company's execution page: C.J. Mahoney, Chief Legal Officer, and Michel Protti, Chief Compliance and Privacy Officer. The state signature pages carry the same date across dozens of offices, from Alabama to Wyoming.

The company concedes nothing. According to the consent judgment, Meta denies the allegations and denies any liability to the plaintiffs. A separate clause states that the agreement does not constitute an admission of liability, wrongdoing, or violation of any local, state, federal, or international law, and that nothing in it establishes a standard of care or serves as precedent in any non-participating United States state or any international jurisdiction. Under Rule 408 of the Federal Rules of Evidence, the document is inadmissible in other proceedings except as needed to approve, interpret, or enforce it.

The claims being released

The states alleged two things. First, that Meta used its technology and its social media platforms to entice, engage, and ultimately ensnare young people, then misled the public about the risks, in violation of state unfair and deceptive acts and practices statutes. Second, that the company violated the Children's Online Privacy Protection Act by collecting, retaining, and using data belonging to residents under the age of thirteen. The states notified the Federal Trade Commission of the action, as COPPA requires.

The release is broad, covering known and unknown claims arising from the conduct alleged. It is also carved up carefully. Criminal liability, securities liability tied to share ownership, tax liability, most antitrust liability, and environmental liability all survive. So do claims by private individuals, and claims by school districts, municipalities, townships, tribes, and other governmental units whose suits are pending in the California coordinated proceeding, in the MDL itself, or in other state and federal courts. A further carve-out preserves claims arising from state investigations into Meta products or business practices identified in a letter from the Colorado Attorney General dated August 25, 2026, the day before entry. That letter is referenced but not attached.

The practical reading is narrow. This judgment closes the state consumer-protection and COPPA front. It does not close the personal-injury front, the school-district front, or whatever the Colorado letter reserves.

What the money actually is

The public figures do not match one another, and the gap is worth setting out precisely.

New Jersey Attorney General Jennifer Davenport, who co-led the multistate effort, announced a settlement of up to $17.1 billion, with a floor of $12.1 billion. According to her office, New Jersey receives at least $525 million and could receive more than $752 million.

Meta's newsroom post describes a payment of approximately $18 billion, with participating states receiving roughly 70 percent, or about $12.7 billion, over the decade, and the remaining 30 percent, or about $5.3 billion, released only if two conditions are met.

Exhibit B to the settlement agreement supports neither round number exactly. It lists, for each of 51 jurisdictions, a Guaranteed Installment Payment and a Contingency Installment Payment, with a footnote clarifying that the figures are per installment rather than totals. The guaranteed installments sum to $1,165,662,174.56 per year. The contingency installments sum to $502,402,600.77 per year. Across ten installments that produces $11.66 billion guaranteed and $5.02 billion contingent, and the exhibit's own total line reads $16,680,647,753.21.

Two further payments sit outside that table. A Cost Fund Payment of $75 million is due within 30 days of the Effective Date, reimbursing state investigatory and litigation costs including National Association of Attorneys General grants. And a Cambridge Settlement Amount of $459,293,017.80, allocated across 48 states in Exhibit E, resolves a separate set of complaints over nonpublic Facebook user information shared with third parties. Add both to Exhibit B and the ceiling reaches $17.21 billion, with a floor of $12.19 billion. Those numbers track the New Jersey release. They do not reach Meta's approximately $18 billion.

New Jersey's guaranteed installment in Exhibit B is $52,564,567.30, which is $525.6 million across ten payments, and its maximum across all installments is $752,199,951.77. Both figures match the state's announcement to the dollar. New York is scheduled for $78,949,551.70 a year guaranteed against a maximum of $1,129,769,577.26. California is the largest line, at $153,594,791.69 a year guaranteed against a maximum of $2,197,944,372.61.

Payment mechanics are fixed. The first guaranteed installment is due within 30 days of the Effective Date, which places it on or before September 26, 2026. Subsequent installments fall on January 15 of each year beginning January 15, 2027. No settling state receives a guaranteed installment until its own consent judgment has been entered; a state whose judgment lands later is paid within 30 days of that entry.

For tax purposes, the states must file a Form 1098-F identifying not less than 50 percent of the amounts paid as compensatory restitution and remediation within the meaning of 26 U.S.C. section 162(f)(2)(A). The Colorado Attorney General's Office files the return on behalf of the group.

Permitted uses are listed rather than mandated. They include expansion of the 988 Suicide and Crisis Lifeline and text-based youth crisis lines, after-school and summer programming, public health advertising credits, a digital wellness public education fund, outdoor activities, youth mental health programming, digital literacy counselors, phone-free school zones, provider training on interactive media use and body dysmorphia, and grants to school districts. California's Exhibit C entry carves $50,000,000 out of each of the first and second annual payments for grants and consumer relief at the attorney general's discretion.

Meta disclosed the accounting consequence itself. According to the company, it expects to accrue a legal expense of approximately $10 billion in the third quarter of 2026 related to the agreement, a charge it says was not contemplated in the expense range given on the second-quarter earnings call, with the other July guidance ranges unchanged. That lands on a business that reported second-quarter advertising revenue of $59.36 billion on July 29, alongside a $2.4 billion legal charge that pushed operating margin down to 31 percent.

The two-hour ceiling

The injunctive core of the agreement is Section II.B, headed Time Management, and it is written in two phases.

Under Phase I, Meta must default teen users to a daily maximum of two hours of use, measured cumulatively across all covered Meta platforms and resetting at midnight in the device's local time zone. Once a teen hits the ceiling, access to the features covered by the limit stops until the reset.

Three categories sit outside the count. Time spent watching Longform Content does not count. Time spent messaging does not count. Time spent in settings does not count, although reaching content through settings does.

The Longform Content carve-out is defined with unusual specificity: video or audio of at least 22 minutes that Meta has determined with a high degree of reliability was not artificially extended. The definition explicitly excludes compilations of shorter material and content padded with a still image or silence to clear the threshold. A teen who exhausts the two-hour allowance on Reels can therefore keep watching a 25-minute video, and the clock does not move.

The agreement anticipates the obvious workaround and closes several doors. When a teen reaches the limit, Meta may not recommend or suggest switching to a different Meta platform, opening messaging, or watching longform video. A neutral screen noting that other parts of the app remain available is permitted, so long as it does not mention messaging. Passive indicators such as an unread-message badge do not count as a recommendation. Messaging and settings functionality must be restricted during the lockout so that a teen cannot reach broader features through them.

Multiple accounts are treated as one. For teens with accounts linked through Meta's Accounts Center or through any explicit linking method, the daily maximum applies cumulatively across all of them. Section II.B.6 goes further, committing Meta to use and keep improving its soft-matching models, naming its Single User Multiple Accounts technology, and drawing on device identifiers, phone numbers, and email addresses to find unlinked accounts belonging to the same teen. Soft-matched accounts also share a single daily allowance. Meta must hand the independent auditor annual data on how well those models perform.

Loosening the default requires a parent. A teen cannot move to a less restrictive limit without approval from a supervising parent, though either party may set something stricter. When a parent does approve a change, Meta must first prompt that parent to choose how long it lasts, offering four options: just for today, one week, 30 days, or until changed back.

Nights, school hours and pauses

Phase I Night Access Mode blocks teens from the covered apps between midnight and 6am local device time. According to the agreement, teens keep messaging and settings access during the block, with the same anti-circumvention restrictions applied. Push notifications are disabled on a wider window, from 10pm to 7am, with an exception for urgent notices tied to account security or platform integrity.

School Mode mutes push notifications during School Hours, which the agreement defines as 8am to 3pm, Monday through Friday, between August 15 and June 15 of every year of the term, subject to adjustment by a supervising parent. Messaging, security, and integrity notifications are exempt. A parent may also restrict all functionality other than messaging during those hours.

Then there are the interruptions. Within four months of the Effective Date, which points to late December 2026, Meta must implement productive pauses by default at 60 minutes and at 90 minutes of cumulative daily use, plus a clear and conspicuous notice after any 15-minute session of continuous use. The agreement requires these to be at least as prominent as illustrative examples supplied in Exhibit G, and to escalate in either length or prominence.

Meta must give the auditor data on how the pauses perform, broken out for teens averaging 60 or more minutes a day and teens averaging 90 or more minutes a day. The company retains discretion over redesign, but is only presumed compliant if it has run a documented assessment, considered changes, and either made and documented them or documented why it declined.

Phase II and the trigger that tightens everything

The second phase is where this agreement stops being about one company.

Phase II applies whenever, during the ten years following the Effective Date, Industry-Wide Adoption has occurred and remains in effect. If it does, the daily limit falls to 60 minutes per app, capped at 120 minutes cumulatively across Meta's platforms, and Night Access Mode widens to 10pm through 7am.

Industry-Wide Adoption is defined against Core Industry Members, which the agreement identifies as Snap, TikTok, and YouTube, for as long as each remains available to United States teens. The trigger fires only when all of them, plus any qualifying new entrant, have either signed a binding commitment with the state in question imposing substantively equivalent obligations, become subject to federal or state law imposing them, or voluntarily implemented and been certified as compliant with them by an independent auditor. There are provisos. Every one of them must also be bound by age assurance requirements for 13 to 17 year olds no less restrictive than Meta's, and each must be subject to at least five years of independent third-party audit of both sets of obligations.

A New SMP Entrant is defined by thresholds rather than names: a product that for four consecutive months lets United States teens create, share, and view user-generated video and images, lets them interact through comments, likes, or subscriptions, is primarily designed for entertainment, social connection, or content discovery competing for the same teen time, and has at least 5 million United States monthly active teens plus an average of at least 30 minutes of daily teen time spent. Products whose primary function is direct messaging, video editing, virtual reality, video gaming, or artificial intelligence chatbots are excluded. The parties record their agreement that no platform outside the three Core Industry Members qualified as of the Effective Date.

Phase I obligations run for five years from the earlier of the Compliance Date or the date Meta implements both the night block and the daily limit. Phase II, if triggered, extends the commitment across the full ten-year term.

The financial architecture is wired to the same trigger. The Contingent Monetary Payment Trigger requires Industry-Wide Adoption on both contingent time management obligations for all Core Industry Members, plus one of two additional conditions: that every Core Industry Member with annual profits above $10 billion is subject to a monetary obligation to that state at least equal to the state's aggregate contingency payments, or that the same holds across a group of settling states. If a state never reaches the trigger during the term, its contingency installments are permanently forfeited and retained by Meta.

That is the structure behind Meta's public appeal to its competitors. According to Mahoney, "this framework will only work if all our peers join us." The company published an open letter on August 26, 2026 urging two named rivals to adopt the same measures. The states have $5.02 billion riding on whether they do.

Age assurance, with numbers attached

Section II.A is the most technically prescriptive part of the document, and it is the part most likely to change how the platform sorts its audience.

Within one year of the Effective Date, Meta must adopt an Age Assurance Framework applying one or more age assurance methods to every covered user in the settling states. Methods fall into two buckets: commercially available ones developed by third parties and licensed to customers, and proprietary ones built by Meta. Any method must be tested annually by an accredited third-party testing provider.

The framework carries hard accuracy targets, expressed as a U18 False Positive Rate, meaning the share of genuine 13 to 17 year olds that Meta incorrectly classifies as 18 or older. For commercially available methods, the thresholds within one year are 10 percent for 16 and 17 year olds and 3 percent for 13 to 15 year olds. Proprietary methods get a longer runway and looser numbers: 14 percent and 7 percent within one year, tightening to 10 percent and 5 percent within two years.

The trade-off is deliberate. Meta is presumptively compliant when it uses a certified commercial method within the vendor's specifications, has not interfered with or undermined it, and has given the testing provider complete information. Choosing a proprietary method forfeits that presumption and adds an obligation of continuous oversight. Testing must cover performance across demographic groups including age, race, gender, and disability status, and for image or video methods must vary photographic conditions including subject presentation, pose, facial archetypes, and lighting. Certification must reference global best practices or ISO 27566 or an equivalent recognised standard.

Circumvention gets its own subsection. Where a user can actively choose a method, such as identity verification or face-based estimation, Meta must cap completed attempts at three per method in any 24-hour period, four per method in any month, and six per method over any two years, counted collectively across a user's hard-linked and soft-matched accounts. A proactive monitoring system must force additional assurance where conduct suggests a user previously assessed as an adult is likely a teen, or a user previously assessed as 13 or older is likely under 13. Anyone who declines the additional check is treated as a teen.

Two provisions carry direct advertising consequences.

The first is the default. For 14 days after account creation, users whose age has not yet been assessed and who state an adult age receive restrictions on contact from suspected adult bad actors and content recommendations inspired by age-appropriate experiences. After 14 days without assessment, they are treated as teen users regardless of stated age, retaining only those two protections. Unresolved age becomes teen status by operation of the agreement.

The second is data handling. Data collected solely for age assurance must be held only as long as needed and then queued for deletion. Meta may retain under-13 data for developing, training, testing, and measuring its detection model, and may retain metadata about which method a user went through where needed for system integrity, deleted within 90 days once no longer required. According to the agreement, under-13 data cannot be used for purposes such as ads targeting and delivery, marketing, or algorithmic optimisation. Stated date of birth, stated age, and the teen-or-adult outcome sit outside those restrictions.

Meta also agrees to incorporate reliable age signals shared by operating systems and app stores run by Apple and Google, with the independent auditor available to make non-binding recommendations on their reliability. That sits alongside a commitment, stated in Meta's own summary of the deal, to keep advocating for legislation requiring app stores to verify age and obtain parental approval before a teen downloads an app. Google's privacy policy director criticised that approach in June 2025, arguing it would expose granular age data to millions of developers and leave desktop and pre-installed apps uncovered.

Finding under-13s

A separate track targets children who should not hold accounts at all.

Within six months of the Effective Date, which points to late February 2027, Meta must maintain or implement several detection methods. Technical measures including soft-matching models must identify other accounts likely belonging to an identified under-13 user and push them into enforcement. In that enforcement process, Meta must assess posts, comments, and other actions for reliable indicia of age, and in the absence of any such indicia must presume the user is under 13 unless there is reliable evidence of a malicious report. A simplified in-app reporting flow, illustrated in Exhibit F, must sit alongside the existing web tool. When an account is removed, Meta must review the deleted user's friend network for further under-13 accounts.

The model commitments run on a longer schedule. Within one year Meta must develop, train, and conduct initial testing of a prototype under-13 age model, reporting to the auditor on its effectiveness against pre-existing methods. At the end of year one it must set a Year 2 Enforcement Target for how many under-13 accounts it will identify and remove, with the auditor reporting to the states on whether that target is reasonable. Within two years the model must be run against all accounts. Missing the target triggers a written explanation and a description of planned model changes. Annual certification of enforced removals follows, audited for methodology.

This is the same gap European regulators are pursuing. The European Commission issued preliminary findings on April 29, 2026 that Instagram and Facebook failed to keep under-13s off the services under the Digital Services Act, with exposure up to 6 percent of global annual turnover.

Feeds, autoplay, likes and filters

Within four months of the Effective Date, Meta must give teens a reasonably accessible option to select a Non-Personalized Feed as their default home feed. The agreement defines that as a feed populated by accounts the teen follows or has friended, in chronological order. Meta keeps the ability to control appearance where reasonably necessary for platform integrity, security, or legal compliance, and the agreement states that such moderation does not count as personalisation.

Prompting is regulated. Within 10 days of identifying a new teen account, and every 90 days after unless the teen disables reminders, Meta must clearly and conspicuously offer the switch. It may not preselect or prioritise either answer, and must present the choice independently of other settings. Parents enrolled in supervision can set the non-personalised feed for their teen, and once they do, reverting requires parental approval.

Autoplay falls under Optional Protective Settings, which must be reasonably accessible, defined as viewable within three user gestures, clearly labelled, easy to notice, visible without scrolling, and discoverable in an intuitive location. Parents can require autoplay to stay off.

Section II.D, headed Social Comparison, contains two sentences with outsized consequences. Meta must disable teens from seeing numbers of likes or reactions by default, on their own posts and on other people's, and that default cannot be changed without a supervising parent. And Meta must disable teens from applying Cosmetic Procedure Filters to their content.

The filter definition draws a line that will require ongoing arbitration. It covers any digital filter or augmented reality effect that distorts, sculpts, redefines, or idealises a face in a way unachievable without cosmetic surgery or extreme makeup technique. It excludes fantasy and character effects turning a user into a non-human, fictional character or animal; makeup and smoothing effects achievable with ordinary technique that do not alter underlying facial structure or meaningfully change skin tone; and parody or exaggeration effects. To operationalise it, the settling states must send Meta illustrative examples and guidance within two months of the Effective Date.

The European Commission's July 10, 2026 preliminary findings on addictive design named infinite scroll, autoplay, push notifications and personalised recommender systems. Three of those four are now addressed by American court order for teens in the settling states, on terms Meta negotiated rather than terms Brussels imposed.

Content standards and a six-hour clock

Section II.E commits Meta to maintain measures it says are already in place: content policies limiting exposure to Age Inappropriate Content, policies against targeting teens with bullying and harassment, interventions interrupting repeated exposure to Sensitive Aggregate Content, defaulting teens into 13-plus content settings, giving parents stricter options, and preventing teens from following, seeing, or interacting with age-inappropriate accounts. For Instagram those are described as current; for Facebook they must be in place by the Compliance Date.

The definitions do real work. Age Inappropriate Content is tied to Meta's Community Standards on bullying and harassment, nudity and sexual activity, child sexual exploitation, sexually explicit language, suicide, self-harm and eating disorders, graphic violence, gambling, and restricted substances, plus policies on high-risk viral challenges and risky stunts. Sensitive Aggregate Content is narrower and more interesting: nutritional discussion, personal weight and fitness accomplishments, muscular physiques, promotion or depiction of non-invasive aesthetic procedures, depiction of body parts commonly associated with body idealisation, and discussion of personal experience living with mental or emotional distress. Individually acceptable, harmful in volume.

Reporting carries a service-level commitment. For potentially harmful reported content submitted in English or Spanish, Meta must maintain processes designed to give teens a decision within six hours in at least 90 percent of cases, with responses required when content is removed, when review finds no violation, or when content is age-gated, and an appeal available where review found no violation.

Section II.E.4 enjoins Meta from making false, misleading, or deceptive representations about the effect or efficacy of teen safety features. That provision applies only to statements made on or after the Effective Date, can be enforced only with the consent of a majority of the six-office bipartisan State Committee, requires 30 days notice and an opportunity to propose corrective action, and is expressly excluded from the independent audit.

Supervision and the parental interface

Parents gain reporting rather than control in most respects, but the reporting is granular. Meta must give supervising parents information on time spent, broken out between the app itself, messaging, and longform content excluded from the daily count, plus usernames of the teen's social connections, of people messaging the teen, and of any user the teen reports. Instagram and Facebook must both notify parents when a teen repeatedly searches for terms related to suicide, self-harm, or eating disorders.

Parents are notified daily when a teen communicates directly with an adult user for the first time, with a link to that adult's profile showing publicly available details such as stated hometown or mutual connections. Where Meta identifies an account as likely to engage in financial blackmail using intimate content, or as likely to engage in inappropriate interactions with children, parents receive a clear and conspicuous notice and resources for discussing it.

Secondary accounts are surfaced. If a supervised teen creates or explicitly links a new account, or is linked to a soft-matched account, the parent is notified with a link to that profile, and the parent's approved time-management settings and content restrictions apply automatically across all of them. Meta must also keep improving controls verifying that a purported supervising parent is actually an adult, using age prediction models and limiting how many accounts one parent can supervise.

The auditor

Accountability sits in Section III. Within 60 days of the Effective Date, pointing to late October 2026, the State Committee and Meta must jointly select a qualified independent third-party auditor to review implementation of the injunctive provisions. Failure to agree routes the question to dispute resolution. The auditor's term begins two months after the Effective Date and runs until 120 days after the fifth final report.

Within 90 days of appointment, the auditor, the committee and Meta must agree a work plan setting out methodology, metrics, scope of information, costs, and a process for Meta to challenge costs as excessive or duplicative. If they cannot agree, the auditor sets a fair and reasonable plan.

Access is broad but bounded. The auditor is entitled to non-privileged information, personnel, systems, and records reasonably relevant to evaluating implementation, including raw and aggregated data, internal documents and communications. Meta may redact privileged material but must describe what was withheld and why. Deadlock escalates to the Chief Privacy and Compliance Officer, then to the State Committee, then to dispute resolution.

Reports run on a four-quarter cycle. The auditor gives Meta a draft within 30 days of a reporting period closing, Meta has 30 days to comment, and the final report follows 30 days after that. Final reports must state whether material gaps or weaknesses were found, the factual basis, and whether they were cured. Meta must prepare corrective action plans within 30 days, secure auditor approval within 90 days, and begin implementation within that same window.

Crucially, the auditor must also publish an executive summary of each final report, excluding nonpublic, proprietary, or confidential material, describing implementation and summarising any recommendations and whether Meta adopted them. Meta pays the auditor's reasonable costs.

Parity clauses that can move the terms

Section V is short and consequential.

The State Monetary Parity Provision applies once Minimum State Participation, defined as at least 40 states or territories, is satisfied. If within 24 months of the Effective Date Meta enters a pre-trial settlement with any other state offering better per-person guaranteed or contingent payments, Meta must pay the settling states the per-person excess multiplied by their combined population, calculated against Census Bureau population estimates for July 1, 2025. Total payments under that clause are capped at the aggregate of all contingency installments.

The State Injunctive Parity Provision is simpler. If within the same 24 months Meta gives another state a lower daily limit, a more restrictive night access mode, or more restrictive school-hours limits, the settling states receive the same term within 30 days of it taking effect elsewhere.

The clause runs the other way too. Under the Meta Injunctive MFN, if a settling state later settles with Snap, TikTok, or YouTube on terms more favourable to that company than Meta's age assurance, time management, or accountability obligations, the state must modify Meta's agreement so that Meta obtains terms at least as relatively favourable. Once such a modification is made and Industry-Wide Adoption is otherwise satisfied, Phase II applies.

The effect is a ratchet in both directions. States cannot go easier on Meta's rivals without loosening Meta's own obligations, and cannot go harder on another state's version of Meta without the settling states inheriting the tougher term.

Where the documents disagree

Five gaps between the executed exhibits and the public announcements are worth recording.

The total. Meta describes approximately $18 billion, split about $12.7 billion guaranteed and about $5.3 billion contingent. Exhibit B totals $16.68 billion, split $11.66 billion and $5.02 billion. Adding the cost fund and the Cambridge payment reaches $17.21 billion. New Jersey's figures of up to $17.1 billion and a $12.1 billion floor sit close to the exhibits. Meta's do not.

The count of jurisdictions. Meta names 52 attorneys general and includes Texas. The settlement agreement's definition of Settling States Attorneys General lists 51 and omits Texas, as do Exhibit B and Exhibit E. Texas appears only in Exhibit D, the list of Eligible States that may still join. Section VI.A.2 provides for amending Exhibit B to add a state that becomes a settling state after the Effective Date, which would be the route. New Jersey's release counts 47 states plus the District of Columbia, Puerto Rico, American Samoa and the Northern Mariana Islands, which is 51. Florida, Guam and the Virgin Islands appear in Exhibit D and nowhere else.

Which rivals are being asked. Meta's open letter names two peers. The settlement agreement defines Core Industry Members as three companies, including Snap. New Jersey's announcement also names Snapchat. Under the agreement's own terms, Industry-Wide Adoption cannot occur unless all three sign.

The Phase II ceiling. Meta describes reducing the daily limit to one hour per app. The agreement sets 60 minutes per platform and adds a cap of 120 minutes cumulatively across platforms, a figure the public summary omits.

The research foundation. Meta's newsroom post describes establishing an independent social media research foundation and sharing consented user data with it. No such foundation appears anywhere in the executed settlement agreement or the consent judgment, including in the definitions, the injunctive provisions, or the exhibits.

The New Jersey release also states that the settlement was approved by the court on the day of its August 26 publication, while Meta's post carries a note that it was updated on August 27, 2026 at 10:30 AM Pacific time to reflect the judge's approval. The signed order is dated August 26.

What changes for advertisers

The mechanics matter more than the money.

Teen inventory contracts on a schedule. A default two-hour cumulative ceiling, a six-hour overnight blackout, notification muting from 10pm to 7am and again through the school day, and escalating interruptions at 15, 60 and 90 minutes together compress the number of teen sessions and the number of impressions inside them. None of this is voluntary product policy that can be quietly reversed; it is a term of a federal judgment enforceable by 51 attorneys general for a decade.

The Longform Content exclusion is the structural detail underneath all of it. Video of 22 minutes or more does not count against the daily limit. Neither does messaging. In a settlement designed to reduce teen time on Instagram and Facebook, the one surface where teen time remains uncapped is long video. That is an explicit incentive to move teen attention toward longer formats, and toward whatever advertising loads those formats carry. Meta has spent the past year pushing Reels deeper into Instagram's ad inventory; the agreement rewards the opposite length.

Audience composition shifts underneath targeting. Accounts unassessed after 14 days are treated as teens regardless of stated age, and false positive thresholds push Meta to reclassify borderline users downward rather than upward. The adult pool available for unrestricted targeting in the settling states gets cleaner and smaller. Meta's age estimation systems already expanded into visual signals in May 2026, and the agreement now attaches audited numerical targets to their accuracy.

Measurement changes for creator and influencer work. Hidden like counts are a default for every teen account, on their own posts and on everyone else's. Campaigns that use visible engagement as social proof toward a teen audience lose that signal at the interface level, whatever the back-end reporting still shows. Cosmetic Procedure Filters, a staple of beauty and cosmetics activation, are disabled for teens entirely, with a definitional boundary that the states will populate with examples within two months.

Geography fragments. The obligations apply in settling states only, and the agreement says expressly that they establish no standard of care and no precedent elsewhere. New Mexico, which litigated separately and secured a $567 million abatement order with a 90-hour monthly ceiling, is not a party. Florida and Texas sit in different positions again. A national teen campaign now runs against at least three different regulatory regimes inside the United States, before European rules are considered.

And the terms can tighten without further litigation. If Snap, TikTok and YouTube settle on comparable terms, the daily limit halves per app, the night block widens by three hours, and the commitment doubles from five years to ten. The states have $5.02 billion of contingency money that only arrives if that happens, which is an unusually direct financial incentive for regulators to keep going. TikTok's own American position moved the other way this month, when the Justice Department moved to vacate its 2019 COPPA decree after a $400 million payment.

Cecilia Kang, the New York Times technology reporter who covered the deal for the paper's daily podcast on August 27, 2026, put the business logic plainly, saying the agreement means "it really strikes at the heart of Meta's business." The episode, published as A Historic Settlement Over Social Media Addiction, frames the deal against the 1998 tobacco settlement rather than against prior technology cases.

Davenport framed it differently. "As a parent, protecting your kids is always your North Star," she said in the New Jersey announcement.

Both framings skip the part that will decide whether the numbers move: enforcement runs through an auditor selected by committee, on a four-quarter reporting cycle, with the first public executive summary still more than a year away.

Timeline

Summary

Who: Meta Platforms, Inc., represented by Chief Legal Officer C.J. Mahoney and Chief Compliance and Privacy Officer Michel Protti, and 51 state, territorial and District of Columbia attorneys general, with New Jersey Attorney General Jennifer Davenport, California Attorney General Rob Bonta, Colorado Attorney General Philip Weiser and Kentucky Attorney General Russell Coleman leading the trial team. Chief United States District Judge Yvonne Gonzalez Rogers entered the judgment. Snap, TikTok and YouTube are named in the agreement as Core Industry Members whose participation would tighten its terms.

What: A consent judgment resolving state unfair and deceptive practices claims and COPPA claims, incorporating a settlement agreement that imposes a default two-hour daily limit on teen use of Instagram and Facebook, a midnight to 6am access block, notification muting from 10pm to 7am and during school hours, productive pauses at 15, 60 and 90 minutes, hidden like counts, disabled cosmetic procedure filters, a non-personalized feed option, an age assurance framework with numerical false positive rate thresholds of 10 percent and 3 percent for commercially available methods, an under-13 detection model with annual enforcement targets, expanded parental notifications, and an independent audit. Exhibit B schedules payments totalling a maximum of $16.68 billion, to which a $75 million cost fund and a $459.29 million Cambridge payment are added, reaching $17.21 billion against a $12.19 billion floor.

When: Meta and the states executed the agreement on August 25, 2026. The joint motion was filed and the judgment entered on August 26, 2026. The Effective Date is today, August 27, 2026. The Compliance Date falls six months later, in late February 2027. Phase I time management obligations run five years; most other obligations run ten years, to August 2036.

Where: The United States District Court for the Northern District of California, in MDL No. 3047 and Case No. 4:23-cv-05448-YGR. The obligations apply to users in the settling states and territories only, and the agreement states expressly that they set no standard of care and no precedent in non-participating states or any international jurisdiction.

Why: The states alleged Meta engineered its platforms to maximise teen engagement while misrepresenting the risks, and collected data from children under 13 in violation of federal law. For the marketing community the consequence is structural rather than reputational. Teen impressions on Instagram and Facebook are now capped by court order across most of the country, long video sits outside the cap, unresolved-age accounts default to teen treatment, engagement counts disappear from teen interfaces, and the whole regime tightens further if three named competitors sign comparable deals.