Governor Gavin Newsom speaking at the signing of California’s child safety bills, 10 September 2026. Credit: Office of Governor Gavin Newsom California has banned personalised recommendation feeds and autoplay for users under 16, in a package of 13 bills signed on Thursday that also imposes crisis protocols, independent audits and annual risk assessments on companion chatbots. Governor Gavin Newsom signed the package on 10 September. The feed provisions prohibit serving minors content ranked on their own history and profile, which is a different instrument from anything else currently in force. “Our children’s safety deserves to be at the center of every conversation about technology.
As innovation moves faster, our protections must keep pace. Today’s legislation makes clear that California will not stand by while unregulated technology puts our children at risk. Innovation comes with responsibility, and protecting our children comes first,” said Governor Gavin Newsom. Australia is consulting on making platforms prompt users to switch ranking off.
Article 38 of the Digital Services Act requires a non-profiling option to exist. California has removed the option and made the answer mandatory for one age group. That distinction is the whole story for anyone watching this from Europe. An opt-out depends on people finding it, which is why Adam Mosseri could tell an Australian briefing this week that engagement falls by half when users switch to a chronological feed: the people who switch are self-selecting.
A ban does not have that problem, and it will produce the first large-scale evidence of what happens to a cohort of teenagers when the ranking is simply turned off. The chatbot half is led by SB 1119, which requires crisis protocols where a user expresses suicidal thoughts, parental controls, notification when a child disables a safety setting, independent child safety audits, and annual risk assessments. It is named Adam’s Law, for Adam Raine. SB 1119 is also one of the four California bills OpenAI publicly backed.
Two days before it was signed, the company asked Congress for mandatory national AI safety rules and described state legislation as a baseline Washington could later codify. This is what that argument looks like when it lands: a law the company supported, with audit and reporting duties attached, in the largest US state. The rest of the package is broader than the headline suggests. SB 867 regulates companion chatbots as toys, which is a product safety framing rather than a speech one and is quietly the most interesting idea in the set.
SB 1276 extends child sexual exploitation offences to AI-generated and digitally altered depictions. AB 1856 requires age verification signals in software applications, AB 2246 governs children’s access to online services, and several bills cover schools, pupil data in AI systems and targeted advertising to children. Age verification is where this collides with European law rather than complementing it. We have reported that Europe’s child safety rules require collecting data its privacy laws forbid, and California’s answer, pushing the signal down into the operating system and the app store, is the one the platforms have been lobbying for.
Meta has argued for a single age check at that layer for two years. It gets it here, in a bill aimed at protecting children. The international pattern is now unmistakable. Canada’s digital safety act also covers chatbots, Australia is legislating on feeds, the US House has passed a kids online safety package, and Roblox has split its users into age tiers ahead of the rules.
Child safety is the route by which platform design regulation is actually passing in every jurisdiction, while general AI and platform bills stall. What the announcement does not give is the part that determines whether any of it works. No effective dates, no penalty amounts, and no enforcement body are specified, and a feed ban for under-16s is only as good as the age assurance underneath it, which is the thing every previous attempt has failed on, as our reporting on age prediction has repeatedly found. California now has the strictest text in the world and the same unsolved problem as everyone else.
For European platforms, the practical consequence arrives regardless. Nobody builds a separate recommendation system for California, so the compliance work done for these rules will show up in products used here, in the same way the state’s privacy law did. Brussels writes the rules Europe argues about, and Sacramento keeps shipping the ones that change the code.
