Today, 26 August 2026, Meta has reached a settlement with a coalition of attorneys general from 47 states and various US territories to conclude the litigation accusing it of designing Instagram and Facebook to addict minors. It will pay at least 12.1 billion dollars over ten years, a figure that could reach 17.1 billion if Snapchat, TikTok, and YouTube accept similar agreements. This is the highest amount a technology company has settled with states for this type of harm. The news will be recounted a thousand times today, almost always focusing on the figure. The significance lies elsewhere.
This is not a conviction: it is a settlement pending judicial approval, and Meta expressly denies having acted unlawfully. It does not admit fault. Anyone writing tonight that 'Meta has been convicted of harming children' will be stating something false, however tempting it may sound. The company is paying to close the case, not because a court has found it liable. That distinction is not a technicality: it is the difference between a proven fact and a pact to avoid proving it.
The fine print details what Meta has agreed to change in exchange for the settlement. What it has accepted is not, primarily, better monitoring of what is published. It is the redesign of the product. Interrupting continuous use with mandatory pauses, setting a two-hour daily limit for minors, blocking access during the early hours, silencing notifications during school hours, hiding 'like' counts, and allowing a feed not governed by the algorithm. Almost all these measures address the mechanics, the attention-capture architecture: the endless scroll, the demanding alert, the comparative metric, the feed that decides for you. Not the content, but the manner in which the content is served.
This is what I have been maintaining in this journal for some time, and for once, it is not just me saying it: it is a multi-billion-dollar agreement. The harm of which these platforms are accused did not lie, or at least not primarily, in specific posts. It lay in the design. In an architecture constructed to maximise time spent, because that time is what is sold to advertisers. When the agreed remedy consists of deactivating infinite scroll and autoplay, turning off nocturnal notifications, and hiding 'likes', it is acknowledging, in fact if not in law, where the problem lay: in the infrastructure, not in the will of the user.
The second shift is the one that truly concerns me. For years, the dominant narrative demanded responsibility from the side that lacked the power. Adolescents were told to use these platforms responsibly, to regulate themselves, to exercise willpower. Parents were told to monitor, to set limits, to speak with their children. A design problem was transformed into a character problem. Yet no individual fourteen-year-old's willpower was competing on equal terms: on the other side were teams of engineers, tests, metrics, and budgets dedicated precisely to overcoming that willpower. Demanding self-control in the face of a system designed to override it is not a solution; it is an alibi. It shifts the blame from the one who builds the trap to the one who falls into it.
What this agreement changes, beyond the money, is the direction of that demand. For the first time at this scale, responsibility is being demanded of the designer, not the user. The minor is not told to 'control your time': the platform is compelled to ensure time has a limit. They are not asked to resist the midnight notification: the notification is switched off. It is a change in the responsible subject, and it is exactly the one I have been advocating for all the infrastructures that surround us, not just these. Individual conduct matters, but it is marginal compared to the decisions made far from the user, in product design, the business model, and now, finally, regulation.
I do not declare victory, for two reasons. The first is that the agreement also mandates the reinforcement of controls regarding dangerous content, harassment, self-harm, and eating disorders, so it would be inaccurate to state that it concerns design alone; what can be stated is that design is its novelty. The second, and most significant, is that correcting Meta does not change the model. As long as the business continues to consist of selling attention, every agreed-upon limit will be a concession regarding a machine that will continue to be built for capture. Twelve billion is a significant sum and, simultaneously, the price of continued operation. Yet, something shifts when redesign ceases to be a request for the company’s good will and becomes an obligation with a figure and a timeline. The question is no longer whether the user will know how to restrain themselves. It is what was built to ensure they could not.
On the open conversation
This text originates from today’s news, Meta’s agreement with state attorneys general, to interpret it through the lens of what I have been maintaining in this series on infrastructures: that responsibility for a system’s harm rests with those who design it, not those who use it. I distinguish the verified fact (it is an agreement, not a conviction; Meta denies wrongdoing; the figure is conditional) from my interpretation. I do not relativise the harm to minors, which I find inadmissible; I point out where it resided and who was improperly asked to answer for it. If anyone wishes to intervene from the fields of law, product design, psychology, or parenting, this notebook remains open.
Sources
Proposed settlement between Meta and 47 states, the District of Columbia and territories (26 August 2026), pending approval (Judge Yvonne Gonzalez Rogers, Northern District of California, Oakland). Statements from the California, New York, Colorado and Tennessee Attorneys General offices.
Coverage: The New York Times, Reuters, CNBC and Courthouse News (26 August 2026). The figure of 'up to 17.1 billion' includes tranches contingent upon settlements with Snapchat, TikTok and YouTube.
2026 precedents: Los Angeles verdict (25 March) and sanctions in New Mexico (375m in March; 942m total following the August order). Meta has announced appeals.
Regulatory context: Australian age-appropriate access legislation (2024, in effect since December 2025); non-legislative resolution of the European Parliament (26 November 2025).