A child is not a mistake in filling a form

16. 6. 2026 / Fabiano Golgo

The UK has announced a ban on social media for children under the age of sixteen. Starting in the spring of 2027, children and teenagers will lose access to at least ten platforms, including TikTok, Snapchat, and YouTube. It is a major, overdue, and understandable decision. For years, governments pretended that it was enough to tell parents to be more attentive, as if an ordinary family could compete with companies that employ engineers to figure out how much longer a finger can keep swiping the screen.

 
The problem is that the British decision is fighting yesterday’s fire. Social media is, of course, still harmful. But children aren’t just looking at the screen anymore. Children are talking to the screen. And the screen responds with a patience that parents and teachers often no longer have by 10 p.m.

That is a different problem. Social media pushes content. Artificial intelligence simulates a conversation. This looks like a technical detail, but it isn’t. A child who watches videos for hours is trapped in an attention machine. A child who confides in a chatbot stands before an answer machine. The first competes for their time. The second may compete for their trust.

That is why it is not enough to debate whether children and teenagers should or should not use AI. They will use it. And in many cases, it can help them: explain a poorly covered topic, translate a sentence, organize an essay, answer a question that no one at home has the time or ability to answer. Banning everything would be like turning a library into a suspicious place just because some shelves aren’t for children.

The serious question is different: why is AI still allowed to pretend it doesn’t know whether it’s talking to a child?

In the physical world, age isn’t a mere detail. A bar can’t sell alcohol to a thirteen-year-old boy just because he claimed at the door that he’s an adult. A movie theatre doesn’t turn a child into an adult by selling them a ticket. But for two decades, the digital world has thrived on this elegant farce: “I’m over thirteen,” “I’m over eighteen,” click, done—a child in pyjamas has become an adult capable of accepting terms of service that even adults don’t read.

The lie was well-known. Children knew it. Parents knew it. Companies knew it best of all. But it served everyone long enough to start looking like a system.

With AI, this lie is more dangerous. Not because the machine is demonic, but because conversation changes the relationship. A chatbot can act like someone. Like a teacher, a friend, an advisor, a confidant. For an adult, this already creates enough confusion. For a child who is just beginning to draw the lines between play, secrecy, authority, and intimacy, the difference between a tool and a companion may be too thin.

The solution isn’t to force every child to show a Silicon Valley ID. That would create a different monster: an industry of verification, biometrics, and document collection sold as child protection. The approach must be smarter and simpler. AI doesn’t need to know the child’s name, address, school, or parents. But it must operate according to age categories.
 
It must know not to speak to a minor as it would an adult.

This requires practical rules.

First: AI services must have a mandatory mode for minors, not parental controls hidden somewhere in a menu that no one can find. This mode should be the default if there is an indication that the user is a child or teenager.

Second: verification should confirm the age category, not the full identity. The system needs to know whether it is dealing with a minor, not to compile a file on them.

Third: minors’ accounts must have limited storage. AI does not need to store a child’s loneliness, fears, and family conflicts in order to “improve the experience.”

Fourth: there must be an absolute ban on romantic and sexual simulations involving minors. This is not “sensitive content.” It is a fundamental civilizational boundary—something we still have to explain using PowerPoint in the digital world.

Fifth: when a child talks about self-harm, abuse, violence, or suicide, the system must exit conversation mode and switch to protection mode. No improvising as a literary therapist. At such a moment, AI doesn’t have to be brilliant. It must be safe, predictable, and steer the child toward real help.

Sixth: these rules must be auditable. It is not enough for a company to publish a manifesto stating that it takes safety seriously. A tech company’s manifesto is often just a funeral mass for accountability. The question is simple: did the system do what it promised, or not?

Seventh: there must be consequences if a company ignores clear signals that it is communicating with minors. Without a substantial fine, everything turns into an awareness campaign with a colourful logo.

Companies will say it’s difficult. Of course it’s difficult. Interestingly, almost nothing is too difficult when it comes to selling ads, predicting behaviour, or keeping a person online for another five minutes. Difficulty only turns into a philosophy when the topic of responsibility comes up.

They’ll also say that kids will find ways around the rules. They will. Kids have always found ways around the rules. They lie about their age, borrow accounts, use their siblings’ phones, and discover paths adults didn’t even know existed. But that has never been a reason to remove boundaries. A minor can get their hands on alcohol, yet a bar is not allowed to sell to children. A teenager can access a restricted movie, yet we do not abolish age ratings.

The law will not eliminate children’s ingenuity. It will eliminate the corporate excuse.

The UK’s decision on social media is important because it challenges the old, comfortable refrain: parents, watch your children more closely. Parents are supposed to watch over them, of course. But parents didn’t design the algorithm, they don’t control the chatbot’s memory, and they don’t know how the system responds to a lonely child at two in the morning. A family cannot be the free security department of a billion-dollar industry.

But if policy stops at social media, it will be regulating the past. The next problem won’t just be how many hours a child spent watching videos. It will be what voice they spoke to, what that voice remembered, what intimacy it offered, and whether it had a duty to know that there wasn’t an adult on the other end.