Chief United States District Judge Yvonne Gonzalez Rogers signed the Meta and State Attorneys General Consent Judgment on August 26, 2026, and it took effect the following day. Within hours the comparison to tobacco was everywhere, most prominently on the New York Times daily podcast, which framed the deal against the 1998 Master Settlement Agreement rather than against any prior technology case.

The comparison is not decoration. The executed documents reproduce the tobacco template with a fidelity that reads as deliberate: a multistate attorney general action settled short of verdict, no admission of liability, a decade of scheduled payments, marketing restrictions in place of a product ban, an industry-participation trigger, and an independent monitor. Four of those five mechanisms carry a documented failure history. Understanding which parts of the tobacco precedent are being copied, and which parts are being quietly left out, is the difference between reading this as an ending and reading it as a starting position.

What the tobacco template actually contained

The Master Settlement Agreement resolved claims by 46 states, the District of Columbia and five territories against the four largest cigarette manufacturers. Four states, including Texas and Florida, had already settled separately for roughly $40bn. The headline number was $206bn over the first 25 years, with payments continuing in perpetuity thereafter.

The money was only part of it. The agreement dissolved the industry's trade and research bodies, forced a document depository that eventually released millions of internal pages, banned cartoon characters in advertising, removed billboards and transit advertising, prohibited brand-name merchandising and sponsorship of youth-attended events, and funded a counter-marketing body that became the American Legacy Foundation. Product regulation arrived separately and eleven years later, when the Family Smoking Prevention and Tobacco Control Act of 2009 gave the Food and Drug Administration authority over the category, including the power to strike the light and mild descriptors that had misled consumers for three decades.

Two features of that history matter most for reading the Meta judgment. The first is that the marketing restrictions bit hard and fast, while the money did comparatively little. The second is that the industry adapted to the measurement rather than to the intent.

The participating-manufacturer trigger, rebuilt

The single closest structural borrowing is the one attracting the least attention.

The Master Settlement Agreement distinguished Original Participating Manufacturers from Subsequent Participating Manufacturers, and contained a Non-Participating Manufacturer adjustment that reduced payments to states if signatories lost market share to companies outside the deal. The clause existed because a settlement covering part of an industry hands a cost advantage to whoever stays out. Its practical effect was to make the states permanent stakeholders in the commercial fortunes of the companies they had just sued.

The Meta agreement inverts the polarity and sharpens the incentive. Payments split into a guaranteed track of $1,165,662,174.56 per installment and a contingency track of $502,402,600.77 per installment across 51 jurisdictions. The contingency money is released only if the Contingent Monetary Payment Trigger fires, and that requires Industry-Wide Adoption across three named Core Industry Members: Snap, TikTok and YouTube. Each must sign binding equivalent obligations, or be captured by equivalent law, or voluntarily implement and be certified compliant by an independent auditor. Each must additionally be bound by age assurance requirements no less restrictive than Meta's and subject to at least five years of third-party audit of both. If a state never reaches the trigger inside the ten-year term, its contingency installments are permanently forfeited and retained by Meta.

That places roughly $5.02bn on the table conditional on the attorneys general successfully prosecuting three further companies. It also explains the open letter Meta published on August 26, 2026, in which chief legal officer C.J. Mahoney wrote that "this framework will only work if all our peers join us." The company is not appealing to solidarity. It is describing the mechanism that decides whether it pays twelve billion dollars or seventeen.

The tobacco version of this clause produced an outcome nobody campaigned for: states with a fiscal stake in the survival of the signatories. The Meta version produces a cleaner alignment, since the states are paid for regulating more rather than for the defendant selling more. Whether that survives contact with three companies who watched what a negotiated deal cost Meta is the open question, and the injunctive parity clause running the other way means any softer settlement with a rival must be back-fitted to Meta's own obligations.

Money without hypothecation

Here the parallel is close enough to be uncomfortable.

The Master Settlement Agreement did not require states to spend tobacco money on tobacco control. They did not. Public health bodies have documented for two decades that states allocate only a small fraction of settlement and excise revenue to prevention and cessation programmes, far below the levels the Centers for Disease Control and Prevention recommends. The rest went to general funds, deficit closure and, in several states, securitised bond issues that converted future payments into immediate cash at a discount.

The Meta agreement repeats the design. Permitted uses are listed rather than mandated: expansion of the 988 Suicide and Crisis Lifeline, text-based youth crisis lines, after-school programming, public health advertising credits, a digital wellness education fund, digital literacy counselors, phone-free school zones, provider training, and grants to school districts. Only California carves out a fixed allocation, reserving $50,000,000 from each of the first two annual payments for grants and consumer relief at the attorney general's discretion. Elsewhere the word is permitted, and the tax treatment requires only that not less than 50 percent of amounts paid be identified as compensatory restitution on a Form 1098-F.

The proportions differ sharply, and not in the direction the headline suggests. Tobacco's annual payments took a double-digit percentage share of domestic cigarette revenue at the time. Meta's guaranteed installment of $1.17bn a year sits against second-quarter 2026 advertising revenue of $59.36bn, an annualised run rate near $237bn. That is under half of one percent. The $10bn legal accrual Meta expects to book in the third quarter of 2026 is a real charge against a real quarter, but it is a one-off, and the recurring obligation is closer to a rounding adjustment than to a structural cost.

Designing to the measurement

The most instructive tobacco precedent is technical rather than legal.

For decades the Federal Trade Commission measured tar and nicotine yields using a smoking machine. Manufacturers responded by perforating filter paper with ventilation holes, which diluted the smoke the machine drew and produced low yields on the test. Human smokers covered the holes with their lips and fingers and compensated by inhaling more deeply. The product measured light and delivered otherwise. It took until 2009 for the descriptors to be prohibited outright.

The Meta agreement anticipates the same behaviour, which is to its authors' credit, and then leaves a channel open anyway.

The daily two-hour ceiling excludes three categories: messaging, settings, and Longform Content, defined as 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 rules out compilations of shorter material and content padded with a still image or silence to clear the threshold. That is a clause written by people who had read the tobacco literature.

It does not close the gap it creates. A teenager who exhausts the allowance on Reels can continue watching a 25-minute video and the clock does not move. In an agreement whose entire purpose is reducing teen time on Instagram and Facebook, the surfaces where teen time remains uncapped are long video and messaging. Meta may not recommend the switch when a teen hits the limit, and a passive unread badge is expressly permitted, but no anti-circumvention clause changes the underlying commercial signal. The agreement rewards length. Meta has spent the past year moving Reels deeper into Instagram's ad load, and the judgment now prices that inventory differently from a 22-minute alternative.

Where the parallel breaks

Three differences are large enough to make the tobacco frame misleading if pushed further.

The first is dose-response. Cigarettes have no beneficial use and a clean, replicated relationship between consumption and mortality. Screen time has neither. The evidence base linking adolescent social media use to psychological harm is correlational, contested in effect size, and confounded by the fact that the same devices carry schoolwork, transport, payment and family contact. A two-hour daily ceiling is therefore an administrative number rather than a threshold derived from harm data, which is why the agreement sets Phase II at 60 minutes per app and 120 minutes cumulative without offering any account of why those figures are the right ones either. The tobacco settlement never had to justify a permitted daily cigarette count, because the answer was zero.

The second is that this agreement writes an acceptable error rate into a federal judgment. The Age Assurance Framework carries a U18 False Positive Rate, the share of genuine 13 to 17 year olds classified as adults. Commercially available methods must reach 10 percent for 16 and 17 year olds and 3 percent for 13 to 15 year olds within one year. Proprietary methods get 14 percent and 7 percent within one year, tightening to 10 percent and 5 percent within two. No tobacco statute ever specified the percentage of minors a retailer was permitted to serve. Probabilistic age estimation at population scale has no counter clerk and no identity document, so the states negotiated a tolerance instead. The countervailing provision is stricter than anything in tobacco law: a user whose age remains unassessed after 14 days is treated as a teen regardless of stated age, which resolves ambiguity against the platform rather than against the child.

The third is that the product and the advertising vehicle are the same object. Tobacco regulation eventually removed the category from broadcast advertising in 1971 and from most European media under the 2003 Tobacco Advertising Directive, and the product carried on being sold in shops. There is no equivalent separation here. Restricting the feed restricts the inventory. Hidden like counts remove a social proof signal that creator campaigns are built on. Disabling Cosmetic Procedure Filters removes a standard beauty activation format outright, with the states due to supply illustrative examples within two months. A non-personalized chronological feed option, offered every 90 days and settable by parents, degrades targeting at the source. These are advertising restrictions imposed through product design rather than through media law, and they arrive without the constitutional argument that has historically constrained commercial speech regulation in the United States.

What the tobacco record predicts

Four things, on the historical evidence.

The marketing and product provisions will do most of the work, and will do it quickly. The MSA's cartoon, billboard and merchandising bans changed youth exposure faster than any payment schedule. The equivalent here is the set of defaults landing at four and six months: productive pauses, the non-personalized feed option, hidden like counts, the filter prohibition, and the overnight block.

The money will mostly not reach the harm. Nothing in the document compels it to, and the tobacco record on that question is unambiguous across every state that received payments.

The disclosure provisions will matter more than their drafting suggests, or not at all. Tobacco's document depository fed thirty years of research and litigation. This agreement's equivalent is an auditor selected within 60 days, reporting on a four-quarter cycle, publishing executive summaries stripped of proprietary material, with the first one more than a year away. Meta's newsroom post described establishing an independent social media research foundation with consented user data. No such foundation appears anywhere in the executed agreement, the consent judgment, the definitions or the exhibits. Whether that gap is closed later is the single best indicator of how seriously to take the transparency limb.

And the category will fragment before it consolidates. The obligations bind settling states only, and the agreement states expressly that they establish no standard of care and no precedent in non-participating states or any international jurisdiction. New Mexico litigated separately to a $567m abatement order with a 90-hour monthly ceiling. Texas appears only in Exhibit D as an eligible state that has not yet joined, alongside Florida, Guam and the Virgin Islands. Meta's own announcement counts 52 attorneys general; the agreement's definitions list 51 and omit Texas. Tobacco took roughly fifteen years to move from a patchwork of state settlements to a single federal regulatory regime under the 2009 Act. There is no equivalent federal statute pending here, and the Justice Department moved in the opposite direction this month when it sought to vacate TikTok's 2019 consent decree after a $400m payment.

The tobacco analogy is sound as far as the machinery goes. Where it fails is the assumption that the machinery worked.