The largest settlement in the history of social media worldwide imposes a two-hour limit on American teenagers, leaves the core mechanism of addiction untouched, and sends the bill to TikTok and YouTube.
At six in the morning on August 26, 2026, federal judge Yvonne Gonzalez Rogers received a document from the parties that effectively wiped out the remainder of a trial expected to last roughly six weeks. Later that same day, the questioning of Instagram chief Adam Mosseri was supposed to continue in the Oakland courtroom. Mark Zuckerberg was next on the state attorneys general’s priority witness list. He never made it to the witness stand.
Meta agreed to pay up to $18 billion and redesign Facebook and Instagram for every American user under the age of 18. A default two-hour daily limit, an overnight feed lockout, hidden like counts, and break reminders every 15 minutes make the list of concessions look substantial. The company admitted no wrongdoing and, under the terms of the settlement, continues to deny every allegation.
The states are celebrating. Civil rights advocates from opposing camps are furious: some say the measures do not go far enough, while others see them as a threat to teenagers’ own rights. I read the Oakland settlement differently. What we are looking at is the most expensive operation in social media history to rescue a business model, and Meta paid precisely to preserve what matters most.
“Don’t Ask, Don’t Tell”: What Menlo Park Knew About Children
When Arturo Bejar’s daughter turned 14 in 2019, he personally helped her set up an Instagram account. She posted about a hobby she shared with her father: restoring vintage cars. Before long, strangers began flooding her direct messages with sexual propositions, photographs of genitals, and requests to show her breasts. The platform responded to most complaints with a standard message: no violation found.
Bejar knew the system from the inside. From 2009 to 2015, he worked at Facebook on cyberbullying, then left to spend more time with his children. For his daughter’s sake, he returned as a consultant to Instagram’s user well-being team and spent two years trying to change things from within. In 2021, he sent senior executives two letters containing statistics on harassment and abuse experienced by users aged 13 to 15, along with examples from his own daughter’s experience. Zuckerberg did not respond. A meeting with Mosseri led nowhere.
In November 2023, Bejar told senators about it, and on August 19, 2026, he told the Oakland jury. His central argument went straight to the top of the pyramid: “Ultimately, it was the corporate culture created by Mark that made it nearly impossible to implement features that would address well-being and safety problems.”
Testimony from former employees should reasonably be weighed with an allowance for personal grievances, but here their claims are backed by documents. Bejar called the heavily promoted “Take a Break” feature doomed to fail because it was not enabled by default, while the company, he said, knew perfectly well that most parents did not have time to dig through settings. He also claimed Meta had tools capable of identifying millions of users under the age of 13 but operated under an unspoken “don’t ask, don’t tell” principle: actively searching for those children would hurt profits.
Former Instagram mental health team employee George Volichenko described his unit’s work even more harshly. According to his testimony, its real function was to protect the company in future litigation. When the team proposed automatically enabling “Take a Break” for younger teenagers, management refused. Volichenko summarized the reasoning this way: “A trade-off against core metrics was undesirable.” The core metrics were how often and how long people used Meta products.
The story is not new. In the fall of 2021, tens of thousands of pages of internal documents taken out of the company by former Facebook product manager Frances Haugen became public. Among them was research showing that one-third of teenage girls dissatisfied with their bodies said Instagram made that dissatisfaction worse. Back in 2019, as part of Project Daisy, Meta tested what would happen if like counts were hidden. Social comparison among teenagers declined, but business metrics declined as well. In the end, Instagram merely allowed users in 2021 to hide likes manually, while Facebook did not adopt the change at all.
The chronology reduces to a simple formula. For years, Meta knew which levers reduced harm, tested them in practice, and chose not to pull them because each one cost money.
Less Than a Month’s Revenue: What an Indulgence Really Costs
The $18 billion figure is staggering only until it is placed beside Meta’s financial statements. In the second quarter of 2026, Meta generated $60.8 billion in revenue, 28 percent more than a year earlier. On average, the company earns about $670 million a day. The entire settlement equals less than 27 days of revenue.
And $18 billion is only the ceiling. About 70 percent, roughly $12.7 billion, will be paid unconditionally, but Meta will spread the payments over ten years: $1.27 billion annually, roughly half a percent of its current annual revenue. The remaining $5.3 billion depends on what competitors do, a structure I will return to later. For perspective, Meta estimates this year’s capital expenditures, largely for artificial intelligence, at $130 billion to $145 billion.
The states initially sought $200 billion. Why did they accept less than one-tenth of that amount? A courtroom victory could have led to years of appeals. North Carolina Attorney General Jeff Jackson explained the coalition’s choice in terms of speed: “We don’t want to lose another generation. We want these changes, and we want them as soon as possible.” It is a substantial argument. A teenager who was 13 in October 2023, when the coalition of states filed its lawsuit, could easily have become eligible to vote by the end of the appellate saga.
Authorities from virtually the entire country joined the settlement, although counts of participating jurisdictions vary from 49 to 52 depending on whether overseas territories are included. Three states remained outside the deal. New Mexico has already won its own case. Texas reached a separate settlement with Meta worth more than $1 billion and containing similar restrictions for teenagers. Florida considered the terms too weak and continues its litigation.
Florida Attorney General James Uthmeier put the objection bluntly: “$18 billion sounds like a huge amount of money, but this is a $1.5 trillion company. It is a rounding error, a few weeks of revenue.” Uthmeier is a Republican and former chief of staff to Governor Ron DeSantis, so there is political calculation in his rhetoric. On the substance, he is right.
Wall Street reads such figures the same way Uthmeier does. In July 2019, the Federal Trade Commission imposed a record $5 billion fine on Facebook over the Cambridge Analytica scandal, and the company’s shares rose after news of the amount emerged. When a New Mexico jury awarded $375 million in penalties against Meta in March 2026, its shares gained 0.8 percent in after-hours trading. Investors learned long ago that litigation costs are simply another operating expense for Meta.
What Meta Gave Up and What It Buried in the Fine Print
The settlement’s central measure is a single two-hour daily limit across Facebook and Instagram for all users under 18, enabled by default. Time spent across every account Meta links to the same teenager will be added together. The restriction can be removed only with parental approval.
Then come the exceptions. Messaging does not count toward the limit, nor does watching video or listening to audio lasting 22 minutes or longer. The cap falls primarily on the feed and Reels. Instagram Direct and Messenger are exempt from part of the restrictions, meaning neither the overnight shutdown nor the two-hour ceiling applies to private communication.
From midnight to six in the morning, teenagers will be unable to scroll the feed, Stories, Explore, or Reels, or publish new content. During school hours, from eight in the morning to three in the afternoon, Meta will disable most push notifications except private messages and account security alerts. Every 15 minutes of continuous use, the system will suggest taking a break. After 60 and 90 minutes of use in a day, it will remind the teenager how much time has been spent. Teenagers will be able to disable autoplay, while parents will be able to prevent them from turning it back on. Like and reaction counts will be hidden by default.
The most important point is buried in what the document does not contain. Meta is not required to disable its personalized recommendation feed. Teenagers will be able to switch to a chronological feed containing posts from accounts they follow. The company will periodically remind them that this option exists, and parents will be able to make that mode mandatory. By default, however, the algorithm remains exactly where it is.
The remaining provisions concern safety. Teenagers will be barred from using filters that simulate plastic surgery. Within a year, Meta must learn to identify children under 13, who are not supposed to be on the platforms at all, as well as teenagers aged 13 to 17 who pretend to be adults, so that teen protections can automatically be applied to them. Parents will be able to see their child’s additional accounts and contacts with suspicious adults.
The most underestimated provision is an independent research fund that Meta will finance and supply with data from users who consent to participate. For the first time, outside researchers will see what until now only the company itself has been able to see.
One telling detail is that Meta is prepared to cut the daily limit to one hour and expand the overnight lockout to 10 p.m. through 7 a.m. if TikTok and YouTube agree to do the same. The generosity is conditional and highly calculated.
A Slot Machine Does Not Shut Off on a Timer
The two-hour limit is the most visible and, in my assessment, the weakest element of the settlement. Screen time stopped being a reliable indicator of harm long ago. A University of Haifa researcher who has studied the relationship between digital media and adolescent mental health since 2010 notes that five hours online was considered pathological back then. Today, it is an ordinary day for a young person. Early smartphone ownership and long hours on social media are associated with depression, eating disorders, and suicidal behavior. Proving causation is far more difficult, and for some teenagers social media can even play a protective role.
A far larger study was conducted at Cornell University. Researchers followed 4,285 American children for four years. At the beginning, their average age was 10. The amount of screen time alone did not predict later suicidal or mental health problems. What did predict them were signs of addictive use: obsessive thoughts about social media, escaping into the feed to avoid problems, failed attempts to stop, and irritability when access was unavailable. Among children whose addictive use intensified with age, the risk of suicidal behavior was roughly 2.1 times higher than among the rest. In the group with consistently high levels, it was 2.4 times higher. By adolescence, roughly one-third of participants fell into groups with either high or increasing levels of addictive use.
A qualification is essential: an observational study identifies an association, not a mechanism. For regulators, however, the implication is clear. Compulsivity is what should be measured. Hours are merely a crude proxy.
Interestingly, Meta itself uses the same criteria. In 2019, its researchers surveyed about 20,000 Facebook users and compared their answers with their actual behavior on the platform. Use was classified as problematic when two conditions were present at the same time: the platform was harming the person’s real life, and the person felt that he or she was losing control. Among American participants, 3.1 percent met the definition. Among users under 25, the proportion was almost twice as high. Instagram alone, according to Zuckerberg, now has 2 billion daily users, and even modest percentages on a base that large translate into tens of millions of people.
The mechanism is also described in the company’s own documents. As early as 2020, Meta employees discussed the “slot machine effect” in internal correspondence. The feed delivers rewards unpredictably. The next video may be worthless, or it may be exactly the one you were hoping for. Likes work the same way. A comment can generate dozens of reactions or none at all. For an adolescent brain, in which the reward system matures earlier than prefrontal control, this kind of roulette is especially powerful. A two-hour timer does not shut it off. It merely closes the casino for the night.
The Algorithm That Finds the Sore Spot
An attorney for Fairplay, an American nonprofit fighting the commercial exploitation of children, identifies the personalized recommendation feed as the primary source of harm. It sharply increases time online and systematically leads children toward progressively more extreme content on the same subject.
Meta’s own internal research from the 2023–2024 school year supports that conclusion. The company analyzed the feeds of 1,149 teenagers. Those who reported dissatisfaction with their bodies were shown roughly three times as much content about eating disorders and body image by the algorithm: 10.5 percent of their recommendations compared with 3.3 percent among their peers. The study did not prove that Instagram caused the dissatisfaction. What it did show was that the platform feeds vulnerable users precisely the material that hurts them. According to the same document, Meta’s automated filters failed to detect 98.5 percent of potentially harmful content.
The picture is the same on TikTok. Amnesty International, together with the Algorithmic Transparency Institute and AI Forensics, created 40 bots designed to imitate 13-year-old users in the United States and Kenya, each displaying different levels of interest in mental health topics. After five to six hours of use, roughly every second video in the feeds of these simulated teenagers involved potentially harmful content about depression, self-harm, or suicide, ten times more often than in the control group.
A 2024 study tested the same phenomenon on real people. Researchers analyzed 1.03 million videos recommended by TikTok to 42 users with eating disorders and 49 members of a control group. The algorithm showed people with eating disorders 146 percent more appearance-related content, 335 percent more material about diets, and 142 percent more fitness content. Toxic content explicitly reinforcing the disorder was shown 4,343 percent more often.
More than 44 times as often. No daily time limit can erase that arithmetic.
The First Amendment as Body Armor for the Algorithm
Why did the recommendation feed survive even Meta’s direct courtroom defeat? New Mexico provided the answer. On March 24, 2026, a jury in Santa Fe found that the company had violated the state’s unfair practices law by misleading the public about the safety of teenagers. The state had sought more than $2 billion. The jury awarded the maximum $5,000 per violation for 37,500 users, a total of $375 million, after deliberating for only one day. Attorney General Raúl Torrez said at the time that Meta executives knew their products were harming children and lied to the public.
In August, Judge Bryan Biedscheid imposed the second part of the sanctions: another $567 million, including $420 million for youth treatment programs, plus five years of judicial oversight of changes to Facebook and Instagram. Meta’s total bill in New Mexico reached $942 million. The attorney general’s office emphasized that, for the first time in the country, a court had rejected the company’s reliance on Section 230 as a shield against liability. Yet the court declined to prohibit recommendation algorithms, infinite scrolling, or autoplay.
The judge cited the First Amendment and the Communications Act of 1934, of which Section 230 became a part. The logic is familiar from Moody v. NetChoice. In July 2024, the U.S. Supreme Court indicated that selecting and ranking content in a feed constitutes a form of editorial judgment protected by freedom of speech. Under American law, an algorithm thus becomes analogous to a newspaper’s editorial policy, and the government cannot dictate to an editorial operation the order in which material must appear.
Supporters of restrictions see this as a category error. The Fairplay attorney argues that a significant portion of the harm comes from design optimized for engagement and that the design can be changed without touching the content of posts. The dispute over where “speech” ends and “mechanism” begins will, in my view, become the central legal battle over social media during the next decade. The settlement allowed Meta to avoid fighting that battle.
In parallel, on March 25, 2026, a Los Angeles jury issued the first verdict in a private lawsuit over social media addiction. Meta and YouTube were ordered to pay $6 million to a 20-year-old plaintiff who began using social media at the age of six. Half of that amount consists of punitive damages. The sum is trivial. The signal is serious: jurors found that the companies acted with malice, oppression, or fraud, and thousands of similar lawsuits are pending in American courts.
Joe Camel Is Back, Now With Reels
In 1988, tobacco company R. J. Reynolds launched Camel advertising featuring the cartoon camel Joe, wearing dark sunglasses and a leather jacket and driving sports cars. Three years later, researchers found that teenagers recognized Joe better than adults did, while Camel’s share of cigarettes smoked by minors had risen from 0.5 percent to 32.8 percent. Under legal pressure, the camel disappeared from advertising in 1997. A year later, 46 states, the District of Columbia, and five territories signed the Master Settlement Agreement with the four largest tobacco companies, at that time the biggest civil settlement in U.S. history: at least $206 billion during the first 25 years alone.
Money was not the main issue. The settlement rewrote the marketing model. It banned advertising aimed at minors, cartoon characters, billboards, branded merchandise, and paid product placement. Cigarettes continued to be sold, and the tobacco giants preserved their market. More than that, the agreement effectively protected established companies from newcomers. Manufacturers that had not joined the settlement were required to place money into special escrow accounts, stripping them of their price advantage.
Comparisons with Meta have become commonplace, but there is a less obvious side to the analogy. A law professor at a Melbourne university explicitly compares the situation with Big Tobacco in the late 1990s: companies agreed to enormous payouts and restrictions if doing so allowed them to preserve the product and their access to consumers. Since December 10, 2025, Australia has prohibited children under 16 from creating accounts on the largest platforms. The American settlement, by contrast, keeps teenagers inside Meta’s ecosystem, preserving brand loyalty and a future adult audience. In effect, this is a warning label on a cigarette pack without a ban on selling cigarettes to children.
The economic logic is similar as well. Tobacco companies desperately needed competitors to be prevented from gaining an advantage by avoiding the same rules. Meta has written that principle directly into the financial structure of its settlement.
$5.3 Billion on a Rival’s Head
Meta will pay the remaining 30 percent of the settlement, about $5.3 billion, only if YouTube and TikTok adopt comparable restrictions for teenagers and themselves reach settlements with the states for corresponding amounts. If they do not, Meta saves billions.
This creates a paradox. From an accounting perspective, Meta benefits if its competitors resist. Market logic says the opposite. If Facebook and Instagram are bound by limits while TikTok and YouTube remain unrestricted, teenagers will simply migrate to the platforms where the feed does not shut down at midnight. According to the largest recurring survey of American teenagers, published in late 2024, about 90 percent used YouTube, roughly 60 percent each used TikTok and Instagram, and only one-third used Facebook. Meta is far from the favorite in this race.
That is why, the very next day after the settlement, Meta bought full-page advertisements in three of the country’s leading newspapers featuring an open letter to TikTok and YouTube. The company called the settlement terms a “new industry standard”: “We want teens to benefit from this new industry standard, but we cannot do it alone. These protections will only be truly effective if we work with our peers, TikTok and YouTube, to implement similar measures.”
The maneuver is brilliant. Overnight, yesterday’s defendant became the author of industry rules. California Attorney General Rob Bonta has already said his office is in contact with Snap, is holding discussions with TikTok, and hopes to bring YouTube to the negotiating table.
Wall Street has already calculated that limits on teenage engagement could hurt YouTube more than Meta itself. TikTok, meanwhile, is now a different company. Since January 22, 2026, its American business has been operated by TikTok USDS Joint Venture, in which Oracle, Silver Lake, and the Emirati investment fund MGX each hold 15 percent, while ByteDance retains 19.9 percent. The new owners, who acquired the asset after years of political drama and threats of a ban, have little reason to seek a war with a coalition of roughly 50 attorneys general.
The biggest surprises lie in the definitions. The standard will apply to new social networks once they reach 5 million American teenagers per month who spend an average of at least 30 minutes a day on them. Explicitly excluded are services whose primary function is private messaging, video editing, virtual reality, gaming, or interaction with artificial intelligence. A mass-market artificial intelligence companion with which teenagers spend two hours a day would not automatically fall under the mechanism. Neither would a teenage social environment in virtual reality. WhatsApp remains outside the perimeter even though it has long offered channels. The founder of Mothers Against Media Addiction calls these exemptions one of the settlement’s biggest flaws.
Now compare those exemptions with Meta’s investment plans. Reality Labs, the division responsible for virtual and augmented reality, lost $4.62 billion in the second quarter of 2026 alone. The company is backing its bet on artificial intelligence with annual capital expenditures of up to $145 billion. Meta accepted strict rules for yesterday’s products while exempting precisely the areas into which it is moving its future. I do not consider that a coincidence.
At the same time, Meta is urging app stores to provide platforms with verified age information, effectively shifting the cost and reputational risk of age verification onto Apple and Google. Utah became the first state in the country to adopt a law based on precisely that model in 2025.
Brussels and Canberra Strike Where Washington Did Not Dare
The day after the American settlement, European Commission spokesperson Thomas Regnier said Meta must provide children in Europe with protections at least as strong as those offered to American children. An investigation into the company under the Digital Services Act has been underway since May 16, 2024. On April 29, 2026, the European Commission preliminarily concluded that the company was failing to keep children under 13 off its platforms, and on July 10 it found that the addictive design of Facebook and Instagram violated the Digital Services Act.
Brussels has directly identified what it believes must change: infinite scrolling and autoplay should be disabled by default, meaningful breaks should be introduced, and recommendation systems should become less focused on engagement. “Protecting the physical and mental health of Europeans must be a priority for social platforms,” European Commission Executive Vice President Henna Virkkunen said on July 10. If the findings are confirmed, the fine could reach 6 percent of global annual revenue. At Meta’s current revenue pace, the ceiling exceeds $14 billion, comparable to the entire American settlement.
The American agreement has energized European regulators. A letter calling for the case to be accelerated was signed by 52 members of the European Parliament. On September 3, Regnier said negotiations were continuing but Meta had not yet offered anything that would allow the case to be closed, and that minor concessions would not be enough. The European Union will seek the disabling of infinite recommendation feeds, autoplay, and push notifications. The difference is fundamental. Brussels is targeting precisely the mechanism that American courts have treated as untouchable.
Australia is going even further. Since December 2025, accounts for children under 16 have been prohibited there, and the eSafety regulator is already threatening platforms with lawsuits for noncompliance. On September 8, Anthony Albanese’s government published draft legislation called “My Feed, My Choice” as part of a broader digital duty-of-care framework. Platforms would be required to notify every user over 16 and offer a choice between an algorithmic feed and a feed containing only content from accounts the user follows. For users under 16, personalized algorithms and infinite scrolling would be disabled by default. Requirements to protect children from addictive design would extend to games, applications, and artificial intelligence chatbots, precisely the areas excluded from the American settlement.
“We will require social platforms to provide tools that give people over 16 a genuine and lasting choice over what they see in their feeds,” Albanese told reporters in Canberra. The weak point of the Australian proposal is money. The maximum fine would be 109.2 million Australian dollars, about $78.6 million. Meta earns that amount in less than three hours.
Parent, Prosecutor, or Teenager: Who Gets the Remote
Not every civil liberties advocate sees the Oakland settlement as a victory. A senior attorney at the Electronic Frontier Foundation, one of America’s oldest organizations defending freedom on the internet, calls it very bad, primarily for teenagers themselves.
The first objection concerns age verification. To apply special rules to children, a platform must first determine who is in front of it. In practice, that means biometrics or identity documents in the hands of a corporation and the end of anonymity for people who may desperately need it, including political dissidents. The Foundation for Individual Rights and Expression warns of another consequence: age-verification systems will spread across the entire internet, and eventually every adult will have to prove his or her age.
The second objection is subtler. The settlement gives parents the authority to decide for teenagers, yet not every child has a safe relationship with his or her parents. The problem becomes especially acute when sexuality, reproductive health, abortion, or medication is involved. State attorneys general will also be able to enforce compliance with age restrictions. California and Kentucky, where abortion is almost entirely prohibited, stood on the same side in Oakland, but their ideas about what information is inappropriate for teenagers differ radically, and the settlement gives each of them leverage over Meta.
“Thirteen-year-olds have rights, and those rights must be respected,” the EFF attorney insists. His prescription is to give teenagers tools that allow them to disable or modify features they themselves consider unsafe.
Child advocates offer the mirror-image response. Fairplay repeatedly hears from families that imposed restrictions and talked to their children about safety, only to find that it was not enough. Parents need baseline standards built into the platform itself. Abandoning social media is not a realistic alternative either. For a teenager, it can mean exclusion from the environment where classmates communicate, reducing the choice to a dilemma between platform risks and social isolation. The ability to assess consequences, consent to data sharing, and resist mechanisms designed to capture attention develops with age, and the law is entitled to take that into account.
The Haifa researcher confronts this dilemma at home. Every time she reminds her 13-year-old daughter about limits, the girl responds that none of her friends have such restrictions. At the same time, the researcher warns against turning the phone into a source of permanent family conflict, because the relationship with parents is itself one of the key factors in adolescent mental health.
I believe both sides are right, and that is precisely why the settlement is poorly designed. It shifts responsibility onto parents and age verification, the most disputed links in the chain, while barely touching the design that extracts attention from every user. Even Fairplay acknowledges that adults constantly ask for the same protections. Infinite scrolling, notifications, popularity counters, and feeds optimized for engagement do not switch off on a person’s eighteenth birthday.
Forecast: Who Is Next Under the Steamroller
My forecast is specific. Judge Gonzalez Rogers will approve the settlement no later than March 31, 2027. The court has no reason to block an agreement supported by nearly every state. TikTok will reach a comparable settlement with the states by June 30, 2027, because its new American owners bought peace, not war. YouTube will hold out longer than anyone else and, absent a courtroom defeat, will sign nothing comparable before the end of 2027. Google has too much to lose, and Wall Street has already done the math. In Europe, Meta will agree by the summer of 2027 to disable autoplay and infinite scrolling by default for minors in order to avoid a formal infringement decision, but it will try to preserve the personalized feed as the default for adults.
One question remains open, and neither the states nor Brussels has an answer to it yet. Who will regulate the feed when it is replaced by an artificial intelligence companion that knows a teenager better than any recommendation algorithm and does not meet the legal definition of a social network?
Arturo Bejar and his daughter restored old cars. That was how her Instagram account began. Any restorer knows that you can repaint a car, reupholster the interior, replace the chrome, install new headlights, and it will still run on its original engine. The Oakland settlement repainted the body. The engine that decides which video a 13-year-old girl sees next is still running at the same speed.