Britain Is Getting Poorer. Our Kids Are Falling Behind on AI.Here is how we fix it — almost for free.

1. A confession before the argument

My son goes to a private school in London. A good one. The kind where the fees would make a working family wince and the buildings have names. And yet, in 2026, he has not been taught a single structured lesson on artificial intelligence. Not what it is. Not how it works. Not how to prompt it. Not how to spot when it is lying to him. Not how to build something small with it on a Saturday afternoon.

I checked with friends whose children are at state schools across London, at grammar schools in the home counties, at famous boarding schools whose alumni run the country. The answer is roughly the same everywhere: a one-off assembly, a worried email to parents about ChatGPT and cheating, a vague policy banning it from homework. That is not a curriculum. That is a panic response dressed up as a policy.

Meanwhile, in Lagos, in Nairobi, in Dhaka, in Manila, in small towns across India that I have visited, children are learning to use AI tools the way my generation learned to use a calculator. Not because their governments are visionary — most are not — but because the tools are free, the phones are cheap, and the kids themselves are hungry. We are about to be lapped by countries we still patronisingly call “developing”.

This essay is not an attack on teachers. Teachers in this country are heroic and exhausted. It is an attack on the absence of political vision, and on the strange British habit of swapping prime ministers and education secretaries every eighteen months as though that were a substitute for actually deciding what we want our children to know.

2. The vision vacuum

Britain is, on most measures that matter, getting poorer. Real wages have been roughly flat for the best part of two decades. Productivity is stuck. Public services creak. We argue about boats and bins while the parts of the world that will eat our lunch are quietly building, training, and investing. Nobody at the top of British politics seems able to say out loud what kind of country we are trying to become in the 2030s, let alone the 2040s.

Education is the place where that vacuum hurts most, because education is the only lever a government has that compounds. Everything else — tax, planning, immigration — is a knob you turn this year for an effect this year. A child taught well at eight pays dividends until they are eighty. A child taught nothing about the technology that will define their working life costs the country for the same span.

And yet there is no national AI curriculum. There is no statutory entitlement for a British child to leave school fluent in the tools that are already rewriting law, medicine, software, design, marketing, logistics, and most of the white collar work their parents do. There is no teacher training programme at scale. There is no examined qualification that signals to an employer, “this young person can think with these machines, not just around them.”

When I ask politely why, the answers I get from people inside the system are honest and depressing. The Department for Education is consumed by exam reform and crumbling concrete. Ofsted is consumed by safeguarding. Schools are consumed by surviving the next budget cycle. AI sits in nobody’s in-tray because it sits in everybody’s. Classic British administrative tragedy.

3. The myth that this is expensive

The single biggest objection I hear, especially from state-school heads, is cost. “We would love to, but we cannot afford laptops, licences, training, infrastructure.” I want to kill this myth gently, because it is the excuse that lets the whole country off the hook.

Teaching a child to use AI well in 2026 does not require a laptop per pupil. It does not require an expensive enterprise licence. It does not require a refurbished computer lab. The marginal cost of getting a British eleven-year-old to a useful working level with modern AI is, conservatively, in the low tens of pounds per year. In many cases it is zero.

Here is why. The frontier models — the ones that genuinely matter — all have free tiers that are vastly more capable than anything that existed in a paid product two years ago. Open source models that run on a single shared classroom machine, or even a decent phone, are now good enough to teach with. The hard part is not access. The hard part is pedagogy, posture, and political will.

4. What “AI literacy” should actually mean

Before we talk about how to deliver it, we have to agree what we are delivering. Too many of the conversations I have with schools collapse because “AI” means ten different things to ten different teachers. Here is the simplest definition I can defend.

A British school leaver in 2030 should be able to do five things without thinking about it. One, hold a useful conversation with a large language model — knowing how to ask, how to push back, how to ask for sources. Two, judge when the model is wrong, biased, or hallucinating, and verify against a primary source. Three, use AI to learn something new — a language, a piece of maths, a historical period — faster than they could alone. Four, build something small with it: a study aid, a revision quiz, a budget tracker, a short film, a song, a piece of code. Five, articulate the ethical and civic stakes — jobs, copyright, surveillance, energy, deepfakes, democracy — in a way that lets them vote and work as informed adults.

Notice what is not on that list. We do not need every child to train a neural network from scratch. We do not need every child to write CUDA kernels. We do not need every child to become a machine-learning engineer, any more than we needed every child in 1995 to become a network administrator. We need fluency, not specialism. Specialism comes later, for the few who choose it.

5. The almost-free plan

Here is what I would do, starting in September, if anyone were asking. None of it requires new primary legislation. Most of it requires courage and a willingness to embarrass the people currently doing nothing.

Step one: a one-page national entitlement

The Department for Education publishes a single page — yes, one page — describing what every British child is entitled to learn about AI by the end of Year 6, Year 9, and Year 11. It is written in plain English. It maps to the five competencies above. It is not a 200-page framework that nobody reads. It is a contract with parents. Cost: effectively zero. Time to draft: a fortnight, if the will exists.

Step two: an open, free curriculum

Commission a public-interest consortium — a couple of universities, the BBC, the Alan Turing Institute, the Royal Society, a few classroom teachers who actually use this stuff — to publish an open-licensed AI curriculum for ages 7 to 18. Lesson plans, worksheets, video explainers, assessment rubrics. Released under Creative Commons. Hosted on a single canonical URL. Any school, anywhere in the country, in the Commonwealth, in the world, can use it. Estimated cost: a few million pounds, one-off. Less than one mile of motorway. Less than one consultancy review of an NHS trust.

Step three: free tools, negotiated centrally

The DfE picks up the phone to OpenAI, Anthropic, Google, Microsoft, Mistral, and the major open-source maintainers and negotiates a national education tier. Every school email address gets a safe, age-appropriate, content-filtered account. The vendors are falling over themselves to do this already — it is cheap marketing for them and it locks in the next generation of users. The state simply has to ask with one voice, instead of leaving 24,000 schools to negotiate alone. Marginal cost to the taxpayer: close to zero. Possibly negative, if structured well.

Step four: train one teacher per school, not all of them

Forget trying to retrain half a million teachers overnight. It will not happen and it is not necessary. Train one “AI lead” per school — an enthusiastic volunteer, paid a modest TLR — through a six-week online programme run by the Open University or similar. They cascade to colleagues. They run an after-school club for keen kids in the first year. They become the in-house expert. Cost: a few thousand pounds per school per year. National total: well under £100 million. For context, that is what we spend on Whitehall consultants in roughly a fortnight.

Step five: a national after-school AI club, free at the point of use

Borrow shamelessly from Code Club and Coder Dojo, both of which have shown that volunteer-led after-school technical education works at scale and at almost no public cost. Set up “AI Club” as a national charity, seeded with a small grant, staffed by volunteers from industry — and there are tens of thousands of engineers, designers, and data scientists in this country who would happily give an hour a week. Run it in libraries, community centres, and the school halls that sit empty after 4pm. Free to every child. Especially the ones whose parents cannot afford a private tutor in this stuff.

Step six: assessment, but lightly

Introduce a short, optional AI-fluency certificate at age 16, modelled on the old CSE or the modern ECDL. Practical, project-based, marked by the teacher and moderated by sample. Not another GCSE. Not another stick to beat schools with. Just a visible signal to employers and to universities that this young person can think with the machines. Cost: minimal, because it piggybacks on existing exam infrastructure.

6. What about the posh schools?

People sometimes assume that the private sector is racing ahead and that the gap to worry about is between independent and state. From the inside, I can tell you it is not. The posh schools are, if anything, more cautious, because they are terrified of a parent complaint, a Times article about cheating, or a safeguarding incident. They are charging £25,000 a year and delivering, on this one crucial subject, roughly the same nothing as the local comprehensive.

The honest truth is that, on AI, the playing field is briefly level. Every British school is at zero. The country that gets this right in the next three years will not be the one with the most money. It will be the one with the most nerve.

7. The third-world mirror

Step seven: an honest conversation about phones

We will not get anywhere on AI in schools while we are still having a moral panic about phones. The current British instinct is to ban the device the AI lives on, then complain that children are not learning to use AI. Pick one. My own view, for what it is worth: phones out of the classroom by default during lessons, phones available in supervised after-school AI sessions, phones used at home with parental guidance. That is not a radical position. It is what most sensible Asian education systems have already settled on. We are arguing about it as though it were 2014.

Step eight: a national children’s AI safety standard

Parents are, rightly, nervous. The single biggest reason schools are paralysed is fear of an incident — a child shown something inappropriate, a chatbot giving harmful advice, a deepfake of a classmate. Most of this is tractable. A short, plain-English standard — what an education-grade AI product must filter, log, age-gate, and disclose — would let heads adopt tools without having to become AI safety experts overnight. The vendors will comply, because the alternative is being shut out of the British market. Cost to the state: the salary of one good civil servant for a year.

I spend time in places British politicians like to pity. In a one-room school outside Lahore last year I watched a twelve-year-old girl use a free chatbot, on a borrowed Android phone, to translate a science textbook from English into Urdu, then quiz herself on it, then ask follow-up questions about photosynthesis that her teacher could not have answered. Nobody had taught her to do this. She had worked it out, because the tool was free and the incentive was overwhelming.

Multiply that girl by a hundred million across South Asia, Africa, and Latin America. That is the cohort our children will be competing with for remote jobs, for university places, for the contracts that used to flow automatically to the West. If we send our kids into that arena AI-illiterate, we are not protecting them. We are disarming them.

8. The political problem, named honestly

None of the above is technically hard. The hard part is that British politics is structurally incapable, right now, of doing anything that pays off after the next election. Education secretaries last, on average, less than two years. AI in schools is at least a five-year programme before the results show up in employment data. No minister currently has the incentive to start it.

This is why I keep coming back, in this civic lab, to the same uncomfortable conclusion. On the long-horizon questions — AI, housing, the NHS, the energy grid, what London is for in 2050 — politicians need to be, as far as possible, taken out of the loop. Not in a sinister technocratic way. In the British way: independent statutory bodies, cross-party compacts, citizens’ assemblies that actually have teeth. The Bank of England model, applied to the bits of national life where the feedback loop is longer than a Parliament.

An independent UK AI-in-Education Authority, with a ten-year mandate, a tiny budget, a small expert board, and a statutory duty to publish an annual report to Parliament on what every British child has actually been taught — that, I would happily campaign for. It is boring. It is unglamorous. It would work.

9. What parents can do on Monday morning

While we wait for the state to wake up, here is what I tell other parents in the school gate conversation. You do not have to wait. Sit with your child for thirty minutes this weekend. Open a free chatbot. Ask it to explain something they are struggling with at school. Then, crucially, ask it to be wrong on purpose, and see if your child can spot the error. Then ask it to help them build something silly — a quiz about their favourite football team, a story in the voice of their dog, a revision plan for next term’s exams.

Thirty minutes a week, for a year, and your child will be ahead of 95% of British school leavers on this. That is a damning sentence about the school system, and a hopeful one about how cheap the fix is.

If you are a teacher reading this, the same applies in your classroom, quietly, regardless of what your trust or your head has or has not decided. The kids will not tell on you. They will thank you in ten years.

10. Closing, and a request

I do not pretend to have all the answers. This is a civic lab, not a manifesto. The plan above is the best one I have been able to sketch on the back of an envelope; I am certain that sharper minds — teachers, researchers, civil servants, parents, and the kids themselves — can improve it.

But the worst thing we can do is what we are doing now, which is nothing dressed up as caution. The UK is getting poorer. Our children are about to inherit that. The single cheapest, fastest, most equitable thing we could give them, this decade, is the ability to think and build with the most powerful tools in human history. It would cost almost nothing. It would change everything.

If you know how to make this happen — in a school, in a borough, in Whitehall, in a charity, in a living room — speak up. The London Model is, in the end, just a suggestion box with a louder voice. Use it.