Four words, from California’s opening: hook, hold, harvest, hide. Infinite scroll, autoplay, push notifications, likes, appearance filters—engineered, the complaint says, against known vulnerabilities of adolescent development, then sold to parents as safe.
A Wednesday, late August, an Oakland courtroom, a federal jury half-settled into its seats, ready to hear witnesses on how Facebook and Instagram might have hooked an entire generation of teenagers into scrolling forever… and Meta Platforms announces something that, a year ago, would have sounded like fiction. Up to $18 billion, paid out over a decade. A cap on how long any identified teen can scroll, across both flagship apps. Notifications muted, midnight to six. Like counts hidden. The cosmetic-surgery filters, gone. No admission of wrongdoing, no apology—just a consent decree, a quiet re-architecture of defaults, and a very expensive exit from a lawsuit filed in 2023 by 29 state attorneys general, later joined by nearly every other state and territory. The states endorsed the deal Thursday; Judge Yvonne Gonzalez Rogers, of the Northern District of California, is expected to approve it. Florida has already walked away, calling the payout “peanuts”. To the casual eye, $18 billion reads as punishment. It isn’t, quite—the number works out to roughly one percent of the company’s projected revenue over the same ten years. A rounding error, dressed as a reckoning.
The money was never the point. It was the opening bid, the thing you ask for so you have something to trade down from. What the states pressed for—what the consent decree actually delivers—is a set of default changes, to how Instagram and Facebook behave for anyone under 18. Two hours a day, cumulative, across both apps and across however many accounts a teenager has quietly spun up. Meta says it will apply the cap even against multiple profiles. Only a parent, a guardian, can lift it.
Midnight to six, the apps go dark. On school days, notifications mute from 8 am to 3 pm—what California AG Rob Bonta called, plainly, “stopping notifications during school”. Fifteen minutes of continuous scrolling triggers a prompt; 60 and 90 minutes of cumulative use trigger warnings. Like counts, hidden by default. Autoplay, optional. Cosmetic-procedure filters, extreme makeup filters—blocked outright, for under-18s.
Here is the distinction that actually carries weight: none of this is opt-in. Instagram has had something like 60 safety tools sitting on the platform for years, and nearly all of them worked on the same quiet assumption—that a teenager, or a tired parent, would go looking for the setting, understand what it did, and switch it on. The new arrangement assumes the opposite. The protective default is on, and turning it off takes a deliberate act. A safety tool a user has to go hunting for, and a safety tool built into the architecture itself.. these are not the same species of thing, whatever the marketing says.
The states didn’t win this on the size of the number they extracted. They won it on paper—35 internal Meta studies, now compiled by NYU Stern’s Tech and Society Lab, many surfaced first by whistleblower Frances Haugen in 2021, and again by Arturo Béjar, a former Instagram safety engineer, in 2023. The BEEF survey, July 2021, overseen by Béjar: 13% of 13-to-15-year-olds reporting unwanted sexual advances, every week. 8% exposed to suicide content, every week. Béjar—who’d returned to Meta in 2019 after his own 14-year-old daughter started fielding repeated unwanted advances on Instagram—later told Congress that “Instagram hosts the largest-scale sexual harassment of teens, to have ever happened”.
In 2019, an internal study called Project Mercury put some Facebook and Instagram users on a forced one-week break; the ones who stopped reported measurable gains in mental health. Leadership, per the documents, discussed how to make those findings go away. A 2019 slide deck: “we make body image issues worse, for one in three teen girls”. A deep dive, same year: “teens have an addict’s narrative, about use”. A senior data scientist, in writing, compared the mechanics to a slot machine—“intermittent rewards are most effective… reinforcing behaviours that become especially hard, to extinguish”.
And then, January 31, 2024, under oath, before the Senate: Mark Zuckerberg saying the existing science “has not shown a causal link, between using social media, and young people having worse mental health outcomes”.
That gap—between what the internal documents already knew, and what the man testified—is the whole case, really. That’s the hinge everything else swings on.
For 30 years, Section 230 has shielded platforms from what their users post. Meta leaned on that shield here too, the way it has everywhere else. But the states didn’t sue over content. They sued over design. Four words, from California’s opening: hook, hold, harvest, hide. Infinite scroll, autoplay, push notifications, likes, appearance filters—engineered, the complaint says, against known vulnerabilities of adolescent development, then sold to parents as safe. That framing is what slipped past 230, and it’s what makes this a precedent-shaped object rather than a mere transaction—the first time a company Meta’s size has answered, even in settlement, for the architecture, not the content sitting inside it.
New Mexico had already extracted close to $1 billion of its own—$567 million into an abatement fund, $375 million in penalties. A Los Angeles jury, in the first individual addiction trial, found Meta and Google liable for $4.2 million to a single teen plaintiff. Neither of those folds into this settlement. Thousands of individual claims remain, with two more trials set for October 28.
Of the $18 billion, $5 billion is conditional—payable only if TikTok, YouTube, and Snapchat adopt comparable protections, and, for TikTok and YouTube, settle their own state claims for at least $5.3 billion apiece. Refuse, and Meta’s number drops to roughly $12 billion. Stanford’s Nora Freeman Engstrom named the mechanism without dressing it up: “Meta is structuring this, to not put itself at a competitive disadvantage”. Carrot for the AGs, cudgel for the rivals, both at once. If it works, a teen capped at two hours on Instagram doesn’t just walk the remaining time over to TikTok. If it doesn’t, Meta banks $5 billion, and its competitors get dragged into negotiations anyway. Full-page ads, in the Times, the Post, the LA Times—an “open letter” to its peers, message unsubtle: this only holds, if everyone signs on.
No legal precedent—consent decrees bind only their signatories. The documents stay partly sealed; as Carolina Rossini put it, “the public lost part of the record, of how we got here”. Trials generate public evidence. Settlements end the generating.
Nothing here resolves the privacy cost of the verification it demands—to enforce an age line, Meta has to know who you are, and a government-ID check pulls in, per one privacy lawyer, “an address, a picture, a variety of other things”, well past a date of birth. Biometrics, vendors, signals borrowed from Apple and Google—undecided, unspecified. The direct-message loophole stays open: the two hour cap and the overnight block don’t reach private chat. A teenager who’s burned her limit on the feed can still talk, all night, to whoever she wants.
Florida stays out—“see them, at trial”, says AG James Uthmeier. And Congress stays out too: the Kids Online Safety Act has sat stalled in the Senate for years, while a coalition of state attorneys general, working consumer-protection law and a decade of patience, did by lawsuit what Washington couldn’t manage by statute.
Béjar, to the BBC, in one line: “At the end of the day, Meta needs to be held accountable, for results, not efforts”.
This is, for now, only an effort. A ten-year decree. An independent auditor for the first five. A research foundation whose governance hasn’t been written yet. The real test arrives in 12, 18 months—when the caps are live, the overnight block is running, the likes are hidden, and someone with the authority and the data actually looks at whether the hours moved, whether the anxiety numbers moved, whether any of it moved at all.
A product’s design is, underneath everything, a set of incentives. Meta spent a decade building ones that rewarded the scroll, rewarded the algorithm, rewarded the staying. The settlement asks it to rebuild those incentives, one default at a time. Whether that holds is the open question—and unlike the settlement itself, no court has ruled on that one yet.
*Brijesh Singh is a senior IPS officer and an author (@brijeshbsingh on X). His latest book on ancient India, “The Cloud Chariot” (Penguin) is out on stands. Views are personal.
