What California's new law gets right, and what it leaves alone
Hello!
On Thursday, Google, Meta and TikTok each filed a lawsuit against California's Attorney General, Rob Bonta, to stop a new state law from coming into force on 1 January. The law is AB 1709. Newsom signed it in September, and it passed both chambers of the California legislature without a single vote against.
I want to spend this whole newsletter on it, because it is the closest thing I have seen to a government taking our argument seriously: that the problem with social media isn't mainly what people post, it's how the platforms are built. It also falls short of where we are, in ways I think are worth being honest about.
What the law actually says
A lot of the coverage calls this an "under-16 social media ban". It isn't, and the difference matters.
The core of it is one sentence: "A covered platform shall not provide an addictive feature to a user who is under 16 years of age." Platforms also have to take "reasonable measures" to make sure under-16s aren't given those features.
So the questions are what counts as an "addictive feature" and what counts as a "covered platform".
Addictive feature. The law defines this as a psychologically exploitative feature meant to maximise engagement that foreseeably leads to compulsive use. It names two things directly:
- the addictive feed: any part of a service that recommends, selects or prioritises posts for you based, even partly, on information tied to you or your device. In plain English, the algorithmic "For You" feed.
- autoplay.
The Attorney General can add more features by regulation. Earlier drafts also named notifications and endless scroll. As far as I can tell, the final text leaves those for the AG to add later, though not every news report agrees on this.
What isn't an addictive feed. This is the part I find most interesting, because it shows how carefully they've drawn the line. A feed is not "addictive" if what drives it is:
- information that isn't persistently tied to you, or about your past interactions with other people's posts
- a search term you've just typed
- your privacy or accessibility settings, or the signal saying you're a minor
- something you explicitly and unambiguously asked for, such as a specific creator, or blocking or deprioritising something (but only if nothing else is feeding into the selection, and nothing autoplays)
- private messages
- the next item in a sequence from the same source (again, without autoplay)
Autoplay stays banned even where the feed itself is exempt.
Covered platform. Any website, app or online service, including but not limited to social media, that offers an addictive feature "as a significant part" of the service. Shopping and review sites are out, as are cloud-storage feeds. Again, the AG can widen this.
How they know you're under 16. This is the clever bit, and the part most critics skip over. The law doesn't ask platforms to collect ID. It uses another California law, the Digital Age Assurance Act, which also starts in January. Under that act, when a phone or computer is set up, the account holder (in practice, often a parent) enters the main user's age, and the operating system passes an age bracket to apps: under 13, 13–15, 16–17 or 18+. The app gets a bracket and nothing else. If that signal isn't available, platforms fall back on an existing standard of having "reasonably determined" the user isn't a minor.
What happens to under-16 accounts. Platforms have two choices: offer under-16s a version with no addictive features, or delete their accounts and associated data. That second option is where the companies' "this is a ban in disguise" argument comes from.
Enforcement. Only the Attorney General or local public prosecutors can sue. There's no private right of action. Fines are up to $50,000 per affected child for a knowing violation, and $25,000 for a negligent one. A court has to take the platform's size and good-faith efforts into account.
What else it does. It sets up a seven-member e-Safety Advisory Commission with members from paediatrics, psychology, tech, education and youth advocacy, including at least one with LGBTQ+ youth policy experience. The commission is advisory only and reports every year.
What it doesn't do. There's no minimum age for having an account and no parental-consent workaround. That's a real change from California's 2024 law (SB 976), which lets under-18s have algorithmic feeds if a parent consents. For under-16s, AB 1709 takes consent off the table. A parent can't sign their 14-year-old up for the algorithm.
The backstory
This fight has already been going on for two years. AB 1709 builds on SB 976, the Protecting Our Kids from Social Media Addiction Act, and the courts have been chewing on that one since 2024.
In short:
- NetChoice, the industry trade group, challenged SB 976. In September 2025 the Ninth Circuit Court of Appeals largely upheld the feed restrictions. It said NetChoice couldn't bring the feed challenge on behalf of all its members, because whether an algorithm is "speech" depends on the platform.
- So Meta, Google/YouTube and TikTok sued individually in November 2025.
- In August this year, Judge Edward Davila refused to block the feed provisions. He found, at least at this preliminary stage, that offering a personalised feed is not an expressive act protected as speech. He then also refused to pause the law while the companies appeal. That appeal is now in front of the Ninth Circuit.
Meanwhile, in March a Los Angeles jury found Meta and YouTube liable for a young woman's mental-health harms from using their platforms as a child. It was the first of the big "social media addiction" trials, and her family argued the companies built their products to be addictive despite their own internal research.
So the companies aren't fighting AB 1709 from a position of strength. They've been losing this argument in California courts, and AB 1709 is a tougher version of a law they've so far failed to stop.
What the companies say
They make broadly the same three arguments.
1. Free speech. The main claim is that choosing which posts to show you, and in what order, is an editorial decision protected by the First Amendment, just like a newspaper choosing its front page. TikTok said curating content based on user preferences is a practice "which bookstores, TV networks and streaming services all use." Meta said the law lets the government dictate how ideas are organised and displayed. Meta's filing goes further: because its feeds are personalised and autoplay by default, it says the law would effectively make it illegal to show minors anything at all, or force it to delete their accounts, unless it changes how its product works.
2. Section 230. All three argue that the federal law shielding platforms from liability for what their users post overrides California's.
3. Parents know best. Google says parents are best placed to make these decisions, points to its existing parental controls, and notes that YouTube autoplay is already off by default for minors and parents can switch off Shorts. It calls the law overly broad. Meta says it would have to either remove the features that keep teens' experience "age-appropriate" or cut thousands of California teens off from their friends and communities.
Outside the companies, the Electronic Frontier Foundation (EFF) opposed the bill and asked Newsom to veto it. Its arguments are that age checks mean more data collected about everyone, adults included, and that a stripped-down version of social media is a de facto ban that cuts young people off from information, community and activism.
What I make of those arguments
I'll try to be fair, because some of this has real substance.
The bookstore comparison doesn't hold. A bookstore decides what goes in the window. It doesn't watch every second you spend looking at each book, build a model of what will keep you in the shop longest, and rearrange the shelves for you personally, in real time, for hours. The feed isn't an editor's judgement about what's worth reading. It's an optimisation loop aimed at your attention. The California courts have so far seen it that way too. I'll admit the Supreme Court's 2024 Moody v. NetChoice decision, which treated content moderation as editorial activity, gives the companies an opening. That's why this will probably end up in front of the Supreme Court eventually.
Section 230 is about the wrong thing. It protects platforms from being treated as the publisher of what other people post. AB 1709 doesn't say anything about what anyone posts. It regulates a delivery mechanism. Under this law, a post that's fine to show in a following feed is equally fine to show in a recommended feed.
"Parents should decide" is the argument that worries me least and convinces me least. Leaving it to parental controls means expecting individual parents to out-manage systems built by thousands of engineers to maximise engagement, on settings the platforms design, default and can change. It's notable that the companies keep pointing to protections they've chosen to switch on, while suing to stop anyone requiring it.
The privacy concern is the strongest, and it's one we share. Age checks that mean uploading your passport or scanning your face to a private company would be a bad trade. AB 1709 is lighter than that: it uses an age bracket set once when a device is set up, and apps only ever see the bracket. So "this means handing over your ID" isn't an accurate description of the law. But it is still age assurance for everyone, adults included, and that's a real cost. I'll come back to this, because it's one of the main reasons we think rules like this should apply to everyone.
The "cut off from community" argument is backwards. Under AB 1709, a 15-year-old can still have an account, follow their friends, message them, search, join communities and post. What goes is the machine choosing what they see next and auto-playing it. If a platform decides it would rather delete teenagers' accounts than offer that, it's telling you which part of the product it actually values.
How this lines up with Unfollow
This is where it gets interesting for us. AB 1709 goes after some of the same features we do, for the same reasons. But it draws its line around age, which is exactly the approach we've argued against.
Where it agrees with us:
- It goes after design, not content. That's the foundation of everything we argue. The harm is in the mechanics, not in individual posts that can be moderated away. California has now written that into law.
- Autoplay is on both lists. It's one of our four asks, alongside public follower counts, infinite scroll and engagement-optimised notifications. California bans it outright for under-16s.
- On algorithmic feeds it goes further than we do. We've argued the default should be a chronological feed of the accounts you follow, with the algorithm there if you opt in. For under-16s, California removes the algorithm completely, with no opt-in and no parental override. As I read the exemptions, a following-only feed with no autoplay survives, which is very close to the default we're asking for.
- It keeps the account and removes the features. We're not against social media, or against following and connecting with people, and neither is this law. Under AB 1709, a 15-year-old keeps their profile, their friends and their messages. What goes is the machinery built to keep them scrolling.
- It could answer our research question. As per our Research page, nobody has directly tested what happens when you remove these features while people carry on using the platforms. If the law survives, from January California becomes that test at enormous scale, and I'd very much like to see it studied properly.
- It shows feature rules can survive the courts. For two years the industry has argued that regulating the feed is censorship, and so far the courts haven't accepted it.
Where it doesn't:
- It stops at 16. This is the big one. We oppose age-based rules because these features don't stop working on people when they turn 16: the French study we cite found two in three adults showed signs of compulsive use. The harm we care about, people scrolling far longer than they meant to at the cost of sleep and time with other people, isn't a childhood problem. Treating it as one leaves the features switched on for most users. See our FAQ for more details.
- It needs age checks for everyone. The device-level age bracket is far lighter than ID uploads, but every phone set up in California will now carry an age declaration, and every covered app has to read it. This both applies a cost we've said age-gating imposes on everyone, and exposes that sensitive data to every app. If the rules applied to all users, nobody would need checking at all and apps would have no reason to have that data on their users.
- Half our list isn't in it. Infinite scroll and engagement-driven notifications were in earlier drafts, but the final text leaves them for the Attorney General to add later. Public follower counts don't appear at all. It's worth noting that the one part of California's 2024 law the appeals court blocked was the rule on hiding like counts from minors. That suggests counts may be harder legal ground than feeds, and something we should think about carefully.
- It draws the boundary in a different place. The law covers any service where a personalised feed is a significant part of the product, so it reaches things we deliberately leave out, like long-form YouTube and Reddit. Our line is about a particular shape of platform: a profile, a follow graph and a constantly updating short-form feed. California's is about whether a machine is picking what you see. Both are defensible; they're just answering different questions.
So I'd call it a welcome half step in our direction. On the features it covers, it's closer to our position than anything else I've seen become law. What we disagree on is who it protects. California accepts that these features are harmful, then protects only teenagers from them. We think the same logic means switching them off by default for everyone, and that once you do, there's no need to check anyone's age.
Why this matters here in the UK
The UK debate is still mostly about age limits and account bans, following Australia's under-16 ban last December, which is already being got around. California has at least picked the better half of the argument: keep the account, and switch off the features built to hook you. That puts the burden on the product rather than the child.
The next step is the one California hasn't taken. If these features are harmful enough that a 15-year-old can't consent to them, and their parent can't consent for them, why should a 17-year-old or a 45-year-old get them by default? That's the case we'll keep making here.
What happens next
The Attorney General's office says it's reviewing the lawsuits and will defend the law. Expect the companies to ask for an injunction before 1 January. In the meantime, the Ninth Circuit's decision on the companies' SB 976 appeal will probably tell us a lot about how this one goes, because the core question is the same: is an algorithmic feed protected speech, or a product design the state can regulate?
I'll keep following it and report back.
As always, if you know people in UK policy, research or campaigning who'd want to talk about any of this, I'd really like to hear from them. Just reply to this email.
If you want more content like this, subscribe here!
Dominic