Loading today’s stories
Product1 distinct publisher3 min readPublished
A former Meta data scientist testified that a proposal to switch Take a Break on by default was refused. About two in every 1,000 teenage users ever turned it on themselves.
The Product Desk · Product desk
Compiled by The Product DeskSomething wrong?How this is made
The number Meta published and the number its own data scientist measured are compatible, and the arithmetic is worth doing once. If more than 90% of teens who set a break reminder keep it on [7], and roughly 0.2% of teenage users ever set one [4], then the promising early result Adam Mosseri described covers about 0.18% of teenage users, fewer than one in 500 [1]. Nothing in the post has to be false for it to be useless. The retention rate was measured on a population the opt-in requirement had already reduced to a rounding error.
Brian Stekloff, cross-examining for Meta, ran the fraction the other way: 0.2% is still hundreds of thousands of teenagers, and Volichenko agreed [10]. True, and it carries an implication the defence may not want said aloud. For 0.2% to reach hundreds of thousands, the weekly teenage user base has to sit somewhere between roughly 100 million and 450 million [2]. The multiplier that makes the covered group sound large makes the uncovered group very large indeed.
The part of the testimony that will travel is the stated reason for the refusal. Volichenko described a lot of fear that an opt-out safety feature would produce large declines in core metrics, and said multiple levels of leadership would then have to report the drop upward and explain why they were tanking the numbers [6]. Read as an operating fact rather than as a scandal, that is a review process in which the cost of a safety default lands on the people who propose it and the benefit lands nowhere that gets reported. No directive is needed to keep a feature opt-in under those conditions.
When he escalated the adoption figures, he was told not to worry, because the team existed partly to protect the company against the coming lawsuits, and his manager told him to announce the findings as improvements to the tool, Carly Nairn reported for Courthouse News [9].
Meta's firmest answer is that the switch has since been flipped: teen accounts carry Take a Break on by default, under-16s need parental consent to turn it off, and the company said in 2025 that 97% of 13 to 15 year olds keep the restrictions [14]. The distance between 0.2% and 97% is a factor of about 485, though the two figures do not measure the same behaviour, one counting teens who went looking for the tool and the other counting teens who left a preset alone [3]. Stekloff also established the ceiling on both numbers: a teenager who opens the app many times a day rather than scrolling in long stretches never trips a session-length reminder [11]. Set either way, this is a settings statistic, not a harm one.
Arturo Bejar, the states' first witness and a former Meta engineer, told the jury the tool was designed to fail, partly because most people never change a setting and partly because the reminder could be snoozed away, and likened it to having to switch on an airbag every time you get into a car [15]. This week's testimony supplies the other half of that argument, in the company's own currency: why the airbag stayed off.
Ranked by verification strength, evidence, and original report placement.
Volichenko testified that roughly two in every 1,000 weekly teenage users took up the excessive-use features, which he called very low, almost negligible.
His team found fewer than 0.165% of teenage users switched the tool on and then took the break, as Madlin Mekelburg reported for Bloomberg; a later review put the figure at 0.2%, which Volichenko called a drop in the ocean.
George Volichenko, a former Meta data scientist, told a jury in Oakland this week that someone inside Instagram proposed switching its break reminder on by default, and that the proposal was rejected because the cost to the app's core metrics was not desirable. Mike Swift reported the testimony for MLex.
Volichenko worked on Instagram's mental wellbeing team from April 2022 to February 2023, and one of his first tasks was to measure how many teenagers used Take a Break, which prompts a 10-minute pause once a session passes a set length.
At the time, Take a Break required users to opt in and was not part of Instagram's default settings.
Volichenko told the jury there was a lot of fear that an opt-out safety feature would produce large declines in core metrics, and that if that happened multiple levels of leadership would have to report it upwards and explain why they were tanking the numbers.
Follow any of these and your For You feed starts watching them — no settings page required.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Sworn testimony relayed through one aggregating outlet
The factual core is sworn trial testimony with specific attributions to courtroom reporters (MLex, Bloomberg, Courthouse News) and named figures, which is strong for a live trial. But the cluster contains a single publisher, no primary transcript or exhibit, and no Meta response to the sharpest allegation; central items remain one witness's account under active cross-examination.
Opt-in take-up negligible; default-on version now shipped
As an opt-in feature the tool reached roughly 0.2% of weekly teenage users, effectively no adoption. The same capability shipped as a default in teen accounts, where Meta reports 97% of 13-to-15-year-olds keep the restrictions on, so real-world presence today is high while user-initiated adoption of the original design was near zero. The score reflects the measured behaviour at issue in the trial, lifted modestly by the confirmed default-on deployment.
Published retention statistic overstated real reach
Meta's public framing that more than 90% of teens keep reminders once set was literally accurate but described fewer than one in 500 teenage users, a substantial overstatement of the safety feature's real reach. The gap is moderated rather than maximal because the article reports counter-evidence: the states' own witness said the communication was accurate, the absolute count is hundreds of thousands of teens, and the default-on version now genuinely covers most teen accounts.
Litigation posture and metric protection on both sides
Incentives are unusually visible and strong: testimony describes leadership fear of reporting metric declines upwards, and a manager allegedly framing the wellbeing team as litigation protection while asking that findings be presented as improvements. The testimony itself is produced inside an adversarial trial with large claimed penalties, and Meta's counsel has a clear incentive to reframe statistics favourably. The witness is subpoenaed and unpaid, which slightly reduces personal incentive concerns.
Specific and internally balanced, but single-source and unadjudicated
Figures, names, dates and reporter attributions are precise, and the article surfaces contradicting cross-examination and Meta's later default change, which supports moderate confidence. Confidence is held below high because there is one publisher, no primary court record, no Meta response on record for the privilege allegation, and the trial outcome is undecided.
invest
The 90% that covered 1.8%: Meta's Take a Break gap hands the states a paper case1 distinct publisher
invest
Meta's Pay Structure, Not Its Policy Page, Is What the States Put on the Stand1 distinct publisher
product
Meta's first witness told a jury the blocker on teen safety was culture, not code1 distinct publisher
product
Meta's under-13 data practices go to a jury: 29 AGs, COPPA, and a porous age gate1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 25, 2026