Meta must cap teen Instagram and Facebook use at 90 hours a month

A district court in Santa Fe fined Meta $567 million on 6 August and imposed a list of conditions on Facebook and Instagram that are meant to run for five years. Together with the $375 million penalty a jury handed down in March, Meta now owes $942 million. Set against a 2025 corporate profit of roughly $60 billion that is not much, and the share price slipped less than half a percent in after-hours trading. What changes in the product is the more interesting part.

90 hours a month, and silence after 10 pm

Judge Bryan Biedscheid imposed a fixed usage cap. Accounts belonging to users under 18 may not be active for more than 90 hours a month, counted across Facebook and Instagram together. That works out to just under three hours a day.

A quiet period comes on top of that. Between 10 pm and 7 am, on every day of the week, Meta may not send push notifications to minors. During the school year the ban also covers 8 am to 3 pm, weekends excluded. Messages from contacts and genuine security or hazard alerts are still allowed.

The third measure covers like counts, the publicly visible number of likes under a post. On accounts belonging to users under 18, Meta has to hide them by default. Turning them back on requires a parent's consent.

The AI chatbots get limits too

The court bars users under 18 in New Mexico from romantic or sexualised conversations with Meta's chatbots, and bars adults there from role play in which a person under 18 appears in a sexualised way. The reasoning is uncomfortable for Meta. The company spent much of the trial explaining how skilfully teenagers get around its own safeguards, then expressed surprise when its chatbots met exactly that behaviour. How firm a grip Meta has on its AI already came up when the company started deleting Instagram accounts automatically.

Why this sets a precedent

Meta had invoked Section 230, the provision in US law that shields platforms from liability for content their users post. The court did not accept it. The state, it held, is not trying to make Meta answer for third-party content but for the effects of its own design decisions. For the first time a platform is on the hook for what it built rather than for what other people put on it.

Algorithms, autoplay and infinite scroll are a different matter. The court ordered nothing on them, even though it calls those features harmful, because Section 230 still protects Meta there. WhatsApp is left out entirely.

What the judge declined to order

Biedscheid refused to mandate real age verification. The Children's Online Privacy Protection Act, or COPPA, bars Meta from collecting personal data from children or tracking them passively, even for the purpose of checking their age. Imposing such a duty on Meta alone would also be inequitable, he wrote. Rules of that kind belong in a legislature rather than a courtroom, and he pointed to the United Kingdom, the European Union and Australia, where parliaments did the work.

The measures apply only to users in New Mexico. How Meta is supposed to identify them is the next open question. Peter Ormerod, a law professor at Villanova University, considers a check against New Mexico IP addresses workable enough. The ruling names no implementation deadline, and Meta has said it will appeal. Separately from all of this, Instagram already lets you decide how much Meta may train on your own photos.