It is true that Meta has a plausible legal strategy for the moment — “served its purpose” and “every case is different” and “must protect the marketplace” — in the narrow sense in which lawyers usually do when they are arguing a multi-front litigation from behind a single well-funded podium. The trouble is that the strategy assumes what the platform’s own architecture disproves: that the addiction is an accident of use rather than a feature of the build.
Consider the symmetry of the position Meta now occupies. It is simultaneously arguing, in court, that its products are not designed to addict young users, and, to its investors, that it has built products so sticky that 3.5 billion people cannot stop using them. It is asking a jury to believe that the thing its entire engineering apparatus has been optimized to do — maximize time-on-screen, maximize engagement-driver notifications, maximize the algorithmic feed that the company’s own internal research showed causes harm to teenagers — is somehow a side effect unrelated to the architecture choice.
The company is planning to spend up to $145 billion on AI infrastructure this year, largely to buy chips and build data centers, while eight thousand laid-off employees pick up their final severance packages. It is asking the market to value it on the future of AI and asking the courts to ignore the past of social media. That is a reasonable ask only if the two things are unrelated — and they are not. The same recommendation-engine architecture that Meta has used to maximize young users’ time-on-screen is the one it intends to scale, under a different name, into the next product cycle. The same data pipeline that feeds the feed feeds the model. The same product-development culture that decided, at every juncture, to prioritize the metric that extracts the most attention from the most vulnerable population is the culture that will decide what the model does.
The four state attorneys general heading to trial in Oakland next month are seeking damages of up to $1.4 trillion. That number, as Meta correctly points out, has no historical analog in consumer-protection enforcement. It is also not an accident that it is roughly nine-tenths of the company’s current market capitalization. The state AGs are not merely asking for a large damages figure — they are calibrating the ask to the enterprise itself, signaling that the harm they allege is existential to Meta’s valuation. And the company, which has already settled at least one case and raised the prospect of altering features in New Mexico, does not need the full sum to land to feel the weight of the precedent building.
The real risk is subtler than the headline damage figure. Meta’s user growth has slowed to a standstill. The state and school-district cases are not only seeking monetary damages — they are asking courts to order Meta to change the features that make its products sticky. If a judge orders default-private accounts for all minors, or consent-required public profiles, or notification blocks during school hours, the company loses the engagement data it uses to train its next generation of AI products. It loses the behavioral pipeline. It loses the foundation of the $145-billion buildout it is currently financing with negative free cash flow.
Meta’s chief privacy and compliance officer, Michel Protti, testified that some of the requested changes had “high potential to grind all our global product development to a halt.” That is a remarkable concession when you stop to think about it. What he is saying — what the architecture of the platform forces him to admit — is that the safety feature the state is asking for is incompatible with the way Meta builds products. The state is asking Meta to stop designing for maximum extraction; Meta is saying it cannot build anything else. The two statements are the same statement, and they amount to an admission that the platform works the way the plaintiffs say it does.
Meta would prefer federal legislation that applies to the whole industry, which is the standard corporate preference for any liability: convert it into a compliance cost that all competitors bear equally, making the rent non-negotiable. The trouble with that strategy here is that the plaintiffs have already won. The New Mexico verdict returned $375 million. The Los Angeles verdict returned $6 million. The precedent those verdicts set — that platform design can be product-defect liability, not content-liability protected by Section 230 — has already started to reshape the legal landscape. The appeal Meta filed in July will test whether that theory survives the Ninth Circuit. If it does, the company will be litigating the same architecture question in dozens of jurisdictions simultaneously, each with its own jury, each with its own timetable, each with its own $1.4-trillion ceiling.
The company’s spokesman said every case is different and the outcome of one doesn’t dictate another. That is true in the narrow sense in which legal spokesmen usually mean it — each venue has its own facts, its own judge, its own jury instructions. It is false in the broader sense that a pattern of outcomes eventually settles into a market expectation. Analysts are already pricing it in. Brian Mulberry of Zacks said the risk is real. Phil Yannella, a data-privacy lawyer not involved in the cases, said no company can ignore the liability. The market has not yet assigned a number to the risk, but it is watching the August trial in Oakland, and it is watching the February trial that follows.
Meta’s AI pivot is not wrong. Large language models are a genuine technological frontier, and the company has the capital, the data, and the talent to compete. The trouble is that the data it intends to train the models on is the same data the lawsuits are trying to make it stop collecting. The subscription product that would replace ad revenue requires user trust to sell. The regulatory environment the company invokes as the proper forum for a solution is the one it has spent the last decade lobbying into paralysis. The company has chosen the most expensive path forward — the $145-billion infrastructure buildout, the layoffs, the litigation defense — without having solved the problem at the architecture level.
The public consultation on the design of social-media platforms is happening in the courts, not at the Federal Trade Commission and not in Congress. The deadline is set by the trial calendar. The jury in Oakland will begin hearing evidence in August. The jury in the consolidated proceeding will take its seat in February. The remedy will be whatever those juries decide is proportionate to the harm. Deadlines are the only part of regulatory processes that the regulated actually respect. Meta has two of them coming.