On Monday, standing outside Downing Street before heading off to the G7, Keir Starmer announced that the UK will ban under-16s from a long list of social media platforms, Snapchat, TikTok, YouTube, Instagram, Facebook and X among them. The line was that we’re “giving children their childhoods back”. The model is essentially Australia’s, which got there first at the end of 2025: platforms that don’t take reasonable steps to keep under-16s out face very large fines, with enforcement aimed at the companies rather than the kids.

I think it’s the right instinct, but I also think it’s going to be a lot messier than Monday’s rhetoric suggested. Like most things in this space, we’re going to have to wait and see.

WHAT IS GOVERNMENT FOR, IF NOT THIS?

Start with the principle, because that’s the part I find hardest to argue with. We already accept that the state restricts access to things we know harm children: alcohol, tobacco, gambling. We don’t pretend a teenager can never get hold of a cigarette; we accept that setting an age limit, putting the legal burden on the seller and changing the default is worth doing anyway. (You could argue there are a few other industries that could do with the same regulatory energy, water companies spring to mind, but that’s a column for another day.)

So “a ban won’t be perfect” was never going to be a knockout argument, and the platforms reaching for it first tells you something. Drink-driving laws aren’t perfect either. The question isn’t whether a measure is leak-proof, it’s whether it shifts the default in the right direction.

THE HARMS ARE REAL, AND THE PLATFORMS MADE THIS BED

The reason I land where I do is simple: this could largely have been avoided if platforms had done a better job of moderation when they had the chance.

I’ve been preparing to move house, which has meant more time than I’d like on Facebook Marketplace, and some of the paid advertising content sitting alongside perfectly normal listings is grotesque, pushing the limits of what’s technically permissible to the absolute extreme, clearly engineered to sit just inside the line.

A few weeks ago, like a lot of people, I was served footage on Instagram of the knife attack in Belfast. No warning, no context, just there in my curated feed. If that’s what an adult who knows how these systems work gets shown unprompted, it’s not hard to imagine what a 13-year-old’s feed looks like.

And it isn’t just the paid stuff. Look at almost any feed now and you’ll find the steady drip of misogynistic content and hate-filled rhetoric that’s become routine, not tucked away in some corner you have to go looking for but pushed to the top, because outrage holds attention better than almost anything else. The same machinery that decided a knife attack belonged in my feed is deciding what a fourteen-year-old boy sees when he opens the app after school.

The research backs this up, and it’s worth being honest about why that matters. Digital harms are hard to study in controlled conditions; you can’t run a clean randomised trial on a child’s developing relationship with an algorithm. But from the evidence we do have, including work the platforms have done on themselves, the picture is clear and damning. Amnesty International’s own research traced how a recommendation feed can pull a young user from mild curiosity about mental health into a rabbit hole of self-harm content in a matter of minutes. That’s not an accident of the system, but rather a system doing exactly what it was optimised to do.

HEAVY ON RHETORIC, LIGHT ON DETAIL

Here’s where the wait-and-see comes in. As our CEO, Guy, described it on Monday, the government’s statement is “heavy on rhetoric and light on detail”, but the detail is where this ban will live or die.

Start with the basics: what even counts as social media? The announcement carves out YouTube Kids and messaging apps like WhatsApp, which is sensible, but the line is going to get tested relentlessly. Is a gaming platform with a chat function social media? What about a live-streaming service? I’ve no doubt we’re about to see a great many grey areas and edge cases designed specifically to probe those definitions (fittingly, much like platforms’ own moderation guidelines…).

While the ban has been being drafted, Meta’s been busy running ads on what feels like every podcast channel going, telling parents about their new “Teen” Instagram accounts with restricted features. Platforms are not waiting for the legislation to catch up; they’re already a few steps ahead, hedging their bets on which direction it goes.

ENFORCEMENT, AND THE LAW OF UNINTENDED CONSEQUENCES

Then there’s enforcement, which is where this gets even messier. Not because a ban is wrong, but because a poorly enforced one could be worse than none at all.

Tech companies move faster than legislators. That’s not a controversial claim, it’s just the shape of the last twenty years. Apple previewed exactly that kind of move a week ago: a suite of new child-safety tools arriving this autumn with iOS 27, including app time limits by category, permission prompts before kids can open a new website and, given what I just described, a Communication Safety feature that will now block gore and violent content in shared images, not just nudity. That could help. But the more the device-makers bake this in, and Google will almost certainly follow on Android, the easier it becomes for the social platforms to point at the phone and claim the problem is handled, sidestepping the harder work of moderating their own products for younger users.

In any case, kids are resourceful, and if enforcement isn’t effective the predictable outcome is that under-16s reach for workarounds, VPNs being the obvious one. The risk there is real and a little perverse: in trying to get around a clumsy ban, a child could be introduced to tools that hand them access to far more harmful corners of the internet than the ones we were trying to protect them from in the first place. A ban that pushes kids towards VPNs and the open web, rather than away from harm, is the failure mode to watch for.

These are manageable problems, and none of them is a reason not to act. But each one is ammunition for a platform that would rather delay and muddy the water than change how its products work.

DELAY AND MUDDY, OR ENGAGE AND SHAPE

Which brings me to the platforms’ choice, because they have one.

We’ve already seen the familiar reflex: come out early, say the ban won’t work, imply the whole thing is naive and the research is flawed. They’ll find plenty of legitimate-sounding hooks to hang that on, whether definitions, enforcement, the VPN problem or children’s rights. Amnesty itself framed Monday’s announcement as “the right diagnosis, wrong prescription”, arguing that the fix for an unsafe-by-design product is to regulate the design, not exclude the user. That’s a serious argument made in good faith, and the platforms will be only too happy to borrow it as cover for doing nothing more than lip service.

With bans now being legislated across Australia, the EU, the UK and beyond, the direction of travel is no longer in doubt, and big changes are coming whether the platforms like it or not. The companies that come out of this best won’t be the ones who spent two years muddying the water; they’ll be the ones who engaged with legislators early and honestly to fix the core issue Amnesty is pointing at, the unsafe-by-design problem, while they still had a say in how.

A NOTE FOR THE PEOPLE WHO BUY THE ADS

Agencies and media buyers are the platforms’ customers; the ad spend is the business. That gives anyone running campaigns at scale a line to their platform reps that most people never get, and a reason to be listened to when they use it.

I recently finished reading Careless People, Sarah Wynn-Williams’ account of her years inside Facebook, and the part that stays with you isn’t the scandals, it’s how ordinary the decisions were that led there: a growth target, a feature shipped, a harm noticed and filed for later. Nobody in the building set out to hurt a fourteen-year-old; the system optimised its way there, one reasonable-looking call at a time.

So, a small ask to start with. Next time you’re on a call with your rep at Meta, Google or TikTok, raise it. Ask what safer-by-design looks like for younger audiences, what they’re building and when.

We’d rather be the company that asked early than the one explaining later why it didn’t. If you’re thinking through what this means for how you reach, and protect, young audiences, get in touch.