Meta deliberately designed Instagram and Facebook to hook teenagers. Four state attorneys general are asking a federal court in Oakland to make the company pay $1.4 trillion for it — nearly the entirety of Meta’s market capitalization — in a trial beginning next month. It is the largest consumer-protection damages demand in American history, and it arrives while the company is spending up to $145 billion this year to rebuild itself around artificial intelligence, having already laid off 8,000 employees to help pay for the pivot.
The architecture designed to keep teenagers scrolling is the thing the company cannot afford to have a jury examine. That is the structural problem underneath the legal problem, and Meta’s own compliance officer has said as much under oath.
It is true that the social-media lawsuits have produced mixed results for plaintiffs. Meta has appealed a verdict in Los Angeles, settled with a Kentucky school district ahead of a planned trial, and seen a teenage plaintiff drop his case without receiving payment. The company says it will continue to fight “baseless” lawsuits. But the theory that is surviving judicial scrutiny and reaching juries is not about harmful content — it is about harmful design, and that distinction cuts through the Section 230 immunity that has insulated social-media platforms from most legal exposure for three decades.
Section 230 of the Communications Decency Act shields platforms from liability for what users post. Judges around the country are now allowing cases to proceed on a different theory: that the platforms themselves — their recommendation algorithms, their notification timing, their infinite scroll, their variable-ratio reinforcement schedules — are products designed to addict. Content is what the user sees. Design is the architecture that determines the user cannot look away. The difference is engineering-substance, not semantics, and it is the reason the Oakland trial matters.
The Oakland trial consolidates claims from the attorneys general of Tennessee, California, Colorado, and Massachusetts. They seek damages of up to $1.4 trillion. Meta has called the number absurd, writing in a court filing that “a sanction of that size has no analog in the history of consumer protection enforcement.” The company’s lawyers are technically correct about the absence of precedent, but the number is not a rounding error — it is the logical endpoint of what happens when you apply a per-violation damages formula to a platform that designed addictive engagement for 3.5 billion users and the enforcement authority happens to belong to state attorneys general rather than private plaintiffs.
The states’ cases carry structural power that personal-injury suits do not. Where individual plaintiffs seek only monetary damages, state attorneys general can ask courts to order Meta to alter its platform features — to change the architecture itself. A separate trial in New Mexico, where a judge has already awarded $375 million in damages, is now weighing a second phase that could reach $3.7 billion. Meta warned investors of that figure in its own filings. The company has signaled willingness to settle some cases, raising the prospect of changes including making all users under 18 private by default and blocking notifications during school hours. These are not hypothetical concessions — Meta’s chief privacy and compliance officer, Michel Protti, testified to them under oath.
But Protti also testified that many of the states’ requests had “high potential to grind all our global product development to a halt.” That sentence is the architecture confessing. The engagement mechanics are not a feature that can be dialed back without consequence — they are the load-bearing structure of the product. Making teenagers’ accounts private by default cuts the network effects. Blocking school-hours notifications interrupts the reinforcement schedule. Limiting algorithmic recommendation degrades the very loop that converts attention into advertising revenue. Meta is telling a court, under oath, that it cannot make its platform safe for teenagers without making its platform less profitable, which is the same thing as saying the platform’s profitability depends on its unsafety.
Meta has pointed to the safety features it has already implemented — teen accounts, stricter parental supervision controls, limits on late-night notifications, compulsory break reminders — as evidence that it prioritizes safety over growth. The company has also argued that court-ordered changes would not stop young people from using other social-media platforms, and that federal legislation applying uniformly across the industry would be preferable to a patchwork of state-by-state rulings. These are not frivolous points. The company is right that federal legislation would be more coherent.
But the argument that uniform federal legislation would be better is an argument about remedy, not innocence, and Meta’s own record on federal legislation is instructive. The company has previously threatened to shut down its platforms in New Mexico rather than comply with state-level mandates — the corporate equivalent of holding its own users hostage — and has argued that many proposed changes are “technically unfeasible.” The safety features Meta promotes are real, but they operate inside an architecture designed around a fundamentally different objective. The break reminder does not undo the infinite scroll. The parental-control toggle does not restructure the recommendation engine. The company is offering safety as a retrofit on a product whose engineering specification is engagement, and the retrofit is subordinate to the specification.
Meta’s user growth has slowed to a standstill. The company’s family of apps added essentially no new users in the first quarter. Analysts expect Meta to report its first quarter of negative free cash flow when earnings come out Wednesday. The company is cutting 8,000 employees — the workforce that built the engagement architecture — to free capital for AI infrastructure it has not yet proven it can monetize. Meta’s chief executive, Mark Zuckerberg, has acknowledged that success in the AI pivot is “not a given.”
The company is planning up to $145 billion in capital spending this year, largely to buy chips and build data centers. It is, in effect, extracting the remaining value from an engagement-based advertising business that has stopped growing to finance a bet on a technology whose commercial returns remain speculative. The extraction is being carried out in the open: laid-off engineers fund GPU clusters, advertising margins cover capital expenditure, and the legal costs of defending the engagement architecture that generated those margins are now competing with the AI budget for the same dollars.
The pattern is legible to anyone who has watched a company hollow out its existing business to finance its next one. The 1995 Gerdau acquisition of Manitoba Rolling Mills followed a version of the same playbook — extract from the existing workforce and plant, invest in the restructuring story, let the workers and communities absorb the cost of the transition. The steelworkers called it something more direct than “capital reallocation.” Meta’s 8,000 laid-off engineers are the workforce analog — discarded to fund a bet whose returns are unproven — and the scale is different, but the structure of the decision is the same: the people who designed the system keep the returns; the people who were designed into the system absorb the losses.
Key social-media lawsuits are now moving toward trials that could reshape how platforms operate — if the verdicts hold. The Oakland trial in August, the February consolidated trial, the New Mexico second phase, and the Los Angeles appeal will collectively determine whether the design-addiction theory survives appellate review and whether state attorneys general can extract structural changes from a platform whose architecture resists them by design.
Meta argues that changing its architecture would harm its global product development. The attorneys general argue that the architecture is the harm. Both positions are defensible, but only one of them is being heard by a jury. The platform that designed engagement as its core product is discovering that engagement has a cost the advertising model priced into growth projections but not into liability reserves — and that the people who can collect on it are not Meta’s engineers but state attorneys general with subpoena power and a theory that survives Section 230.
The inheritance of extraction eventually arrives at the courthouse door. The architecture was always designed to keep teenagers scrolling. The question the Oakland trial will answer is whether a company can be made to pay for having designed it that way — and whether $1.4 trillion is the right number for a product that monetized adolescent attention as a feature, not a bug. The engagement architecture Meta built to extract attention from teenagers has met the one force it cannot twiddle: a courtroom.