A default with a parent-only override is a different piece of software from a settings toggle. To raise the two-hour daily ceiling on an under-18 account, the system has to know which adult is the parent and be confident the account belongs to a minor in the first place [2]. That is why the same document obliges Meta to work harder at identifying under-18 accounts, remove under-13s, and submit the practice to an independent auditor [15]. Age inference stops being a growth heuristic nobody outside the company can inspect and becomes an audited control.
The money is structured the same way, as leverage rather than a lump. Meta pays 70 percent over ten years into states' online safety programmes [7], which is $12.6bn, averaging about $1.26bn a year [1]. The remaining tranche is conditional: Meta hands it over only if YouTube and TikTok between them match the amount and adopt a one-hour cap, a nighttime block and age assurance [8]. Note that 30 percent of $18bn is $5.4bn, about $100m more than the figure attached to the tranche [2], so the "up to" in the headline number is doing real work.
Meta's own framing is that the protections "will only be truly effective if we work with our peers", naming TikTok and YouTube [6]. Read as a spec rather than a press line: Meta has offered to halve the cap to one hour and stretch the dark window from six hours to nine [5][4], but only if rivals sign. It has published its strictest defaults and declined to ship them alone, which tells you roughly what it believes those defaults cost in engagement.
The interruption budget is now arithmetic. Prompts land every 15 minutes of continuous use, with cumulative reminders at 60 and 90 minutes [4]: at most eight continuous-use prompts inside a two-hour ceiling, four inside the one-hour version [3]. Around them sit four more mechanism changes aimed at the same loop, like and reaction counts hidden from teens [9], extreme makeup filters banned alongside the existing cosmetic surgery block [10], a non-algorithmic feed a parent can set as the default [11], and an autoplay off-switch parents can also pre-set [12]. These are close to an itemised reply to the attorneys general, who alleged Meta deployed "harmful and psychologically manipulative product features" to induce compulsive use while telling the public those features were safe [17].
Two limits on the read. A court still has to approve the proposal, and Meta denied wrongdoing in it [1], so the delivery dates that matter to anyone building a competing product do not exist yet. And the settlement landed mid-trial, before Instagram head Adam Mosseri was due back on the stand [19], which is not the posture of a company that expected to argue these numbers to a jury. The terms above come from a single published account, duplicated across the material supplied here [5].
