Adam Mosseri, who has run Instagram since 2018, testified on Tuesday at the federal courthouse in Oakland, California, in the second week of the trial brought against Meta by 29 US states, Al Jazeera reports.

He said take-up of the Take a Break feature, which prompts users to close the app after long sessions, had been "in the low single digits", reaching one to two percent of accounts before Instagram made it standard in Teen Accounts, launched in 2024.

The sentence both sides will use

"Most teens didn't want it. We decided to push forward with it anyway."

For Meta, that is evidence of good faith: the company built a protection users did not ask for and eventually imposed it by default. For the states, the same sentence describes a product whose optional safeguards were used by almost nobody for years while the company knew it.

Both readings are available from the same words, which is why we are quoting them in full rather than characterizing them.

What is actually being alleged

Four states, led by California, Colorado, Kentucky and New Jersey, claim Meta deliberately designed its platforms to addict young users, causing anxiety, depression and suicide risk. All 29 allege improper collection of data from children under 13. Damages sought are around $200 billion.

Meta denies the addiction claims and says research has shown no clear connection between adolescent social media use and reduced wellbeing.

That is a genuine scientific dispute rather than a rhetorical one. The literature on adolescent mental health and social media is contested among researchers, with disagreement about effect sizes and about causation running in either direction. A court will have to decide it on the evidence presented, which is a different exercise from settling it scientifically.

The design question underneath

The one-to-two percent figure is the most useful number in the story, because it is not about harm at all. It measures how many people used a protection that had to be switched on.

Whatever a court concludes about addiction, that finding will be familiar to anyone who has studied defaults: optional safeguards are used by almost no one, which is why the shift to default settings in 2024 is the substantive change and the years before it are the exposure.

What we could not establish

We could not establish when the trial began or is expected to end, which states brought which claims in detail, what other testimony has been heard, or the judge's name. We verified this account from a single publication.