Product1 publisher2 min readPublished
California requires a separate unranked feed for every user under 16
Governor Gavin Newsom signed 13 child safety bills on 10 September, and the feed rules bar serving under-16s content ranked on their own history and profile. Europe's DSA requires only that a non-profiling option exist.
The Product Desk · Product desk
What happened
- Governor Gavin Newsom signed a package of 13 child safety bills on 10 September, and one of them bans personalised recommendation feeds and autoplay for users under 16.
- Article 38 of the Digital Services Act requires only that a non-profiling option exist for European users, leaving the choice with the person using the app.
- SB 1119, named Adam's Law for Adam Raine, requires crisis protocols where a user expresses suicidal thoughts, parental controls, and notification when a child disables a safety setting.
- AB 1856 requires age verification signals in software applications, and other bills in the package cover schools, pupil data in AI systems and targeted advertising to children.
- The announcement carries no effective dates, no penalty amounts and no named enforcement body.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- decision The compliance artefact has moved from the settings row to the ranking layer. Someone has to own a ranking path that stays on, and own the rule that decides which accounts land in it.
- constraint TNW's read is that platforms run one recommendation system for everywhere, so the under-16 path built for Sacramento becomes the one European teenagers get.
- exposure OpenAI's argument for federal rules now has audit and annual risk-assessment duties attached to a bill it endorsed, in the largest US state.
- capability Regulators get a new measurement: a large teenage cohort with ranking off, put there by the rule instead of by going to find the setting.
An opt-out is a settings screen. One ranking service, a flag on the account, a chronological branch behind the flag. California's feed provisions prohibit serving minors content ranked on their own history and profile [3], so the under-16 account needs a ranking path it stays in, plus a routing decision made at signup by an age signal the platform trusts enough to act on. Australia is consulting on making platforms prompt users to switch ranking off [4].
The number product teams will reach for is Adam Mosseri's. He told an Australian briefing this week that engagement falls by half when users switch to a chronological feed [7]. TNW makes the right objection to using it here: the people who switch are self-selecting [8]. The figure describes users who went looking in settings and found the option. A 14-year-old routed to the unranked feed at signup only ever sees that one.
The accuracy of the whole thing is inherited. AB 1856 requires age verification signals in software applications [15], which pushes the check down into the operating system and the app store [16]. That is the layer Meta has spent two years arguing should carry a single age check [17]. TNW has also reported that Europe's child safety rules require collecting data its privacy laws forbid [18].
The chatbot half is where the lobbying shows. OpenAI publicly backed four of the bills [12], under a third of the 13 [1][1], and two days before signing it asked Congress for mandatory national AI safety rules, describing state legislation as a baseline Washington could later codify [13]. SB 1119 also carries independent child safety audits and annual risk assessments [11]. SB 867 regulates companion chatbots as toys, a product safety framing rather than a speech one [14]. "Innovation comes with responsibility, and protecting our children comes first," said Governor Gavin Newsom [21].
What you build depends on whether you can route on an age signal someone else generated, and on whether an unranked feed can fill a first session for an account with no history behind it. Both true, and the work is plumbing and a QA matrix. Signal good and feed empty, and the problem is cold start: regional popularity, follows, human curation. Feed good and signal weak, and you misroute in both directions, and the misroute that gets escalated is the 17-year-old dropped into the under-16 path. Neither true, and what ships is the toggle California removed [6].
What to watch
- Whether platforms fence the unranked under-16 path to California or route every minor account into it worldwide.
- Whether any published retention data for the mandated unranked cohort resembles the halving Mosseri described for self-selected switchers.
- Whether OpenAI cites Adam's Law as the baseline in its next request for federal AI safety rules.