Ofcom Tests the All-Seeing Machine That Says It Can Spot a Child

Ofcom is investigating TikTok’s age-assurance system. Britain’s online-safety regime has finally reached the question every platform would rather turn into fog.

This piece is satirical commentary drawing on the sources listed below.

Well Leg’ends here we are again,

Ofcom has opened a formal investigation into TikTok.

Not a consultation, not a warning letter dressed as progress, not another glossy promise from the “safety and innovation” cupboard.

An actual investigation.

On 16 July 2026, Ofcom announced that it would examine whether TikTok Information Technologies UK Limited has failed, or is failing, to comply with certain child-protection duties under section 12 of the Online Safety Act 2023.

There it is. The thing Britain has spent years promising, debating and gently lowering into the warm bath of regulatory language. A regulator is now asking whether one of the largest attention machines on Earth can do the first job every child-safety regime depends on.

Can TikTok reliably tell when a child is using TikTok?

No breach has been established. That must be said clearly, because law is not vibes in a novelty wig. Ofcom has stated that opening an investigation does not mean it has reached a conclusion about TikTok’s compliance. TikTok says it complies with the law and uses expert-informed technology to detect users’ ages and protect younger users.

That is the fair position.

It is also not the end of the matter.

Because the question now moves past the safety brochure, past the polished statement, past the carefully upholstered phrase “ongoing commitment”, and into the machinery itself.

TikTok is not a passive noticeboard. It is not a corkboard in a village hall where someone has pinned a Zumba flyer next to a lost cat and a damp announcement about bridge club.

Its entire product is inference.

It watches what users watch, what they skip and where they hesitate. It learns the pause, the replay, the second glance. It works out what will hold attention before the user has fully admitted the interest to themselves.

It can detect a developing fascination with air fryers, skincare routines, conspiracy clips, protein powder, performative cleaning, men pressure-washing driveways and seventeen different forms of emotional collapse set to dance music.

It can look at a thumb movement and decide, “Yes, this person is now interested in sourdough, celebrity divorce and a man in Ohio explaining the fall of civilisation with a whiteboard.”

Yet the approximate age of the person holding the telephone remains, apparently, one of technology’s great regulatory pressure points.

The all-seeing machine develops the visual acuity of a mole in sunglasses.

Ofcom says evidence in its latest age-assurance work reinforces serious doubts about some age-inference models used by social-media companies. In some cases, those models may have failed to detect significant numbers of children correctly, potentially exposing them to harmful material. TikTok’s approach has now become the subject of a formal investigation.

That does not prove failure.

It does prove that the regulator has stopped admiring the lock on the leaflet and started checking whether the lock is actually on the door.

That is the absurdity sitting in the middle of modern platform regulation. The companies sell themselves as supernatural engines of personalisation. They know your mood, your weakness and your bedtime. They know when you are bored, when you are suggestible and when your brain has gone soft enough to accept a raccoon washing grapes as emotional support.

But when the law asks, “Is this user a child?”, Ofcom’s concern is that the machine’s age-inference tools may not be strong enough. The birthday box is not the whole system, but it is still the front door, and the regulator is now checking whether the lock behind it works.

TikTok’s stated minimum age is 13. That number matters, but the problem has never been the number printed in the terms and conditions. The problem is whether the platform can distinguish between the age somebody declares and the age they are likely to be.

A minimum age without effective enforcement is not a barrier.

It is a small sign beside an open gate.

It is a bouncer made of cardboard.

It is a child-safety regime asking the child to mark its own homework.

Under section 12 of the Online Safety Act, regulated user-to-user services likely to be accessed by children must use proportionate systems and processes to prevent children from encountering specified categories of harmful content. The relevant child-safety duties came into effect on 25 July 2025.

Ofcom says social-media companies must use highly effective age assurance to identify child users and tailor protections accordingly. It has also said that some companies rely mainly on age inference, even though age inference is not listed in Ofcom’s guidance as a method capable of being highly effective for this purpose.

There is the hinge.

There is the little metal part that decides whether the door opens properly or comes off in your hand.

The government can announce restrictions. Parliament can create duties. Ministers can go on breakfast television promising to restore childhood to its pre-smartphone condition while standing beside a school fence and doing the grave face.

None of it works unless somebody can identify the children.

That is the first decision. Everything else rests on top of it.

Get that wrong and the child is treated as an adult, the safety tool goes to the wrong shelf, and the feed carries on humming like a fridge in a betting shop.

This is where the policy stops being abstract. It becomes a child alone with a phone, a parent cooking tea in the next room, a school dealing with the fallout on Monday morning, and a company explaining afterwards that its systems are always improving.

There is the human consequence. Not in the press release. Not in the ministerial quote. Not in the safety centre with pastel icons and rounded buttons.

On the phone. In the hand. During the scroll.

That is where the gate either works or it does not.

And this is not only about what TikTok hosts. It is about what TikTok recommends, repeats and pushes back into the hand after the first swipe has already happened.

There is a difference between harmful material existing somewhere on a vast platform and a platform’s systems helping users find it. One is a warehouse problem. The other is a sales assistant with your browsing history.

In May 2026, Ofcom said TikTok and YouTube had not committed to significant changes to reduce harmful material being served to children, maintaining that their feeds were already safe. Ofcom said its evidence suggested otherwise.

Ofcom’s research found that 73 per cent of 11- to 17-year-olds encountered harmful content during a four-week period. Just over a third of those children recalled encountering it while scrolling through a feed. Among secondary-school-aged children who saw harmful material, 53 per cent recalled seeing it on TikTok, more than on any other platform listed by Ofcom.

Those figures do not prove that TikTok breached the law. They do not prove that every exposure was caused by TikTok’s systems. They do explain why the regulator is interested in the relationship between age detection, product design and recommendation.

Seventy-three per cent. That is the number. Not a vibe, not a moral panic in a tabloid hat, not a committee hallucination after too much instant coffee. Children are encountering harmful material, feeds are part of the route, and companies that sell miraculous personalisation cannot expect regulators to accept vague age-detection confidence as enough.

So test the machine.

Not the brochure.

The machine.

That is exactly the evidential question Ofcom says it will now examine.

Because moderation after the event is only one part of the story. A platform can remove individual videos while leaving intact the machinery that repeatedly pushes similar material towards the same kind of user. It can sweep the floor while the conveyor belt keeps dropping glass.

The Online Safety Act recognises this. It requires providers to consider how algorithms may affect children’s exposure to harmful material. In plain English, the law understands that harm can come not only from what exists, but from how the system distributes it.

That is where the real fight lives.

Inside the feed, inside the ranking, inside the recommendation loop that says, “Here, have another.”

Ofcom is not merely asking whether TikTok deletes prohibited material after somebody reports it. It is examining whether the systems around the material are effective enough to keep it away from children in the first place.

That is a harder question.

It walks past reception, ignores the branded water bottles, opens the staff door and goes looking for the engine.

Ofcom deserves credit for opening the inquiry. That sentence may feel strange. Do not panic. Nobody has joined a fan club. No one is knitting a regulatory scarf.

Credit where it is due: the regulator has picked up the file.

But an investigation is not a finding. A finding is not enforcement. Enforcement is not automatically success.

The Online Safety Act will not ultimately be judged by the number of investigations announced, letters issued or webpages created with stern headings and calm fonts. It will be judged by whether the conduct and design of major platforms change when the regulator identifies a failure.

Ofcom says its first step will be to use formal powers to gather and analyse evidence. It expects that process to take at least three months and intends to issue an update in October 2026.

If the regulator eventually concludes that TikTok has breached its duties, it can impose a fine of up to £18 million or 10 per cent of qualifying worldwide revenue, whichever is greater. In the most serious cases, it may ask a court to require payment providers, advertisers or internet service providers to withdraw services or block access in the UK.

Those powers look formidable on paper.

So does a gym membership in January.

The question is what happens when the paper meets one of the largest technology platforms in the world.

Will Ofcom require meaningful changes to TikTok’s age-assurance system? Will it examine how recommendation works for accounts believed to belong to children? Will it demand evidence that changes work, rather than accepting another sofa-cushion sentence from the compliance department, plumped, scented and placed carefully over the stain?

This is where online safety stops being a slogan and becomes a stress test.

A statutory hammer is useful only if someone is prepared to swing it. Otherwise it is just a shiny object in the cupboard, next to the emergency guidance and the packet of biscuits reserved for stakeholder engagement.

The warning signs were already there.

This is not TikTok’s first difficult encounter with Ofcom over child-safety information. In July 2024, the regulator fined TikTok £1.875 million for providing inaccurate information about its Family Pairing parental-controls feature. Ofcom found that TikTok’s data-governance processes had been inadequate and that the company had been slow to alert the regulator to inaccuracies.

That case arose under the earlier video-sharing-platform regime, not the present section 12 investigation. The 2024 fine does not prove anything about the current investigation. It does, however, sit in the background as a useful reminder that accurate evidence from major platforms is not a decorative extra. It is the foundation on which regulation depends.

If the regulator cannot trust the information it receives, enforcement becomes theatre with a spreadsheet.

Then came the Global Witness testing in October 2025.

Global Witness reported that it conducted seven tests involving TikTok accounts registered with the birth date of a 13-year-old. It said investigators enabled Restricted Mode but were nevertheless shown sexualised search suggestions and encountered pornographic material during every test. In one case, Global Witness reported that pornographic material appeared after the first suggested search term was selected.

Those were Global Witness findings, not Ofcom findings. Again, the distinction matters.

But the questions were obvious.

If an account is registered as belonging to a 13-year-old, and Restricted Mode is enabled, what should happen next? Should the system treat that account as a child account in a meaningful way? Should search suggestions and recommendation behave as though a child is present?

Or should the safety layer sit there like a laminated sign in a chip shop window, technically visible, practically irrelevant?

Global Witness said its tests were conducted both before and after the Online Safety Act’s child-safety duties came into effect on 25 July 2025. It shared the findings with Ofcom and called for a formal investigation.

Ofcom has now opened one.

That does not mean every allegation has been proved. It means the question has moved from external testing into formal regulatory examination.

The file has entered the room.

TikTok says it complies with the law. It says it uses expert-informed technology to detect users’ ages. It points to safety measures including private-by-default teen accounts, restrictions on direct messaging for users under 16 and enhanced age-assurance technology.

That response must be included because fairness is not optional.

It must also be tested because reassurance is not performance.

The question is not whether TikTok has safety features. The question is whether those safety features work well enough for the duties Parliament created.

A policy is not a shield if the system leaks around it. A setting is not a safeguard if the feed keeps finding a way through. A safety feature is not a child-protection regime just because the word “safety” has been given a badge and a lanyard.

This is the difference between presentation and performance.

Between a lock and a sticker of a lock.

Between “we take this very seriously” and “the child did not encounter the material”.

That is the only sentence that ultimately matters.

The child did not encounter the material.

That is where policy has to land. On the phone. In the bedroom. At the bus stop. During the scroll nobody is supervising because parents are human beings, not 24-hour content moderators with a kettle and a mortgage.

On 3 October 2025, SPN published “Mandatory Enforcement and Systemic Failure: TikTok vs. UK Law”, with the subtitle “The Online Safety Act Meets Its First Stress Test”.

The article argued that the real test of the legislation would not be whether TikTok had safety policies, Restricted Mode or reassuring language. It would be whether age-assurance and recommendation systems actually kept harmful material away from children.

It called for Ofcom to open a formal investigation into those systems.

Ofcom is now investigating that fault line.

That does not mean the earlier article has been proved in every respect. Some of its language treated alleged statutory breaches more categorically than the current regulatory position permits. Today’s announcement begins an evidential process. It does not complete one.

But the central diagnosis has travelled. The issue has moved from external testing and editorial argument into formal regulatory examination.

That matters.

Because the question was never whether TikTok could produce a safety policy. Of course it could. Any company of that size can produce a safety policy. There are probably entire cupboards of them. They breed in ring binders. They reproduce by PDF.

The question was whether the systems worked.

Whether the age gate was a gate. Whether Restricted Mode restricted. Whether the recommendation machinery understood that a child is not just an engagement opportunity with a schoolbag.

And whether the Online Safety Act had teeth, or whether Britain had built a large, expensive regulatory scarecrow and dressed it in a hi-vis jacket.

Now we find out.

Britain has spent years debating whether children should be protected online, as though there were a serious political movement demanding that childhood be fed into a content blender for quarterly growth.

There is not.

The real argument has always been enforcement.

Who must act? What must change? How do we know it changed?

For years, the platform economy has benefited from the gap between technical complexity and public patience. The systems are complicated. The interfaces are friendly. The harms are dispersed across millions of screens. Responsibility becomes mist. Everyone is concerned. Nobody is quite holding the mop.

Parents are told to supervise. Schools are told to educate. Children are told to be resilient. Regulators are told to be proportionate. Platforms are told to innovate responsibly.

And somewhere in the middle of that laminated nonsense, an 11-year-old is scrolling through a feed designed by adults, optimised for attention, monetised at scale and explained afterwards in language so soft you could lose a tooth in it.

Enough.

The point of law is to move responsibility back up the chain.

Not to the child. Not to the exhausted parent trying to understand seventeen privacy settings after work. Not to the teacher dealing with the fallout on Monday morning.

To the company that designed the system. To the company that profits from the system. To the company that claims the system can understand users with astonishing precision until the moment public duty knocks on the door.

That is why Ofcom’s investigation matters beyond TikTok.

Every proposed restriction on children’s social-media use depends on the same basic capability.

The system must recognise a child before it can decide what that child should be allowed to see.

If it cannot do that, the whole thing becomes pantomime regulation. A velvet rope across an open field. A ministerial announcement wearing shoes two sizes too big. A locked gate painted on cardboard while the real gate swings open behind it.

Ofcom has picked up the statutory hammer.

Good.

Now comes the harder part.

Use it properly.

Do not just count policies. Test outcomes.

Do not just review settings. Examine behaviour.

Do not just ask whether the company has a process. Ask whether the process protects the child.

That is the line.

That is the whole bloody line.

Because the public does not need another page of platform reassurance. It does not need another sentence beginning “we are committed”. It does not need another compliance sofa with the cushions arranged to hide the springs.

It needs proof that the machinery works.

Proof that children are recognised as children. Proof that harmful material is not merely removed after the damage. Proof that systems designed for attention can be forced to respect duty.

TikTok says its systems comply with the law.

Ofcom will now test that position.

That is how this is supposed to work.

Not by vibes, not by panic, and not by another minister doing grave-face beside a school fence while the machine keeps serving children whatever keeps them watching.

Evidence. Process. Finding. Enforcement, if warranted.

That is the adult sequence.

And yes, it is slower than a headline. It is less satisfying than instant outrage. It does not fit neatly into the hot little toaster slot of online argument.

But if Britain is serious about online safety, this is where seriousness begins.

Inside the age-assurance system. Inside the recommendation loop. Inside the machine that knows everything except, allegedly, whether the person holding the phone is a child.

Nine months ago, we asked whether Ofcom would pick up the statutory hammer.

It has.

Now we find out whether Britain built an enforcement regime, or merely commissioned an impressive collection of tools and left them in a cupboard marked “future consideration”.

The children do not need another cupboard.

They need the gate to work.

Further reading: “Mandatory Enforcement and Systemic Failure: TikTok vs. UK Law: The Online Safety Act Meets Its First Stress Test”, published 3 October 2025.

Stay warm, stay loud, and stay deeply, gloriously allergic to bullshit.

Willy & Bill

---

SPN is free to read, but not free to produce. If this helps you make sense of the circus, support the chaos.

We’re at 782 Leg’Ends. Help us reach 1,000.

218 more Leg’Ends gets us to 1,000. If one person sent this to one politically exhausted friend, we’d be there by breakfast.

💸 Go paid if you can.
☕ Buy me a coffee if you fancy a one-off thank you.
If money’s tight, share, like, restack, or comment. The paying Leg’Ends have you covered.

This piece draws exclusively on publicly available documents, official statements, parliamentary records, and verified investigative reporting. Satirical interpretation is applied at the level of tone, metaphor, and narrative framing only. No factual claims have been altered for effect. The satire lies in the exposure of the absurdity, not in invention.

Receipts / Reading

Right, here is the paper trail. This is the evidence locker. The argument in the piece is satirical. The underlying factual claims are drawn from the sources below.

---

TikTok Investigation

https://www.ofcom.org.uk/online-safety/protecting-children/investigation-into-tiktoks-compliance-with-duties-to-protect-children-from-encountering-harmful-content-under-section-12

Age Assurance

https://www.ofcom.org.uk/online-safety/protecting-children/age-checks-helping-make-online-experiences-safer-for-uk-children-but-job-not-done-and-tech-industry-must-act-to-strengthen-protections

Age Report

https://www.ofcom.org.uk/online-safety/protecting-children/use-of-age-assurance-report-2026

Unsafe Feeds

https://www.ofcom.org.uk/online-safety/protecting-children/tech-firms-commit-to-stronger-anti-grooming-measures-in-response-to-ofcom-demands

Children’s Research

https://www.ofcom.org.uk/online-safety/protecting-children/younger-phone-owners-the-rise-of-ai-and-consumption-over-creation-our-latest-look-at-uk-childrens-media-lives

Global Witness

https://globalwitness.org/en/campaigns/digital-threats/tiktok-directs-13-year-olds-to-porn/

Previous Fine

https://www.ofcom.org.uk/online-safety/protecting-children/tiktok-fined-1.875m-for-providing-inaccurate-data-on-safety-controls

Ofcom Guidance

https://www.ofcom.org.uk/online-safety/illegal-and-harmful-content/guide-for-services

Regulation Roadmap

https://www.ofcom.org.uk/online-safety/illegal-and-harmful-content/roadmap-to-regulation