An MP asked the Department for Education what it was doing to improve AI skills. The answer, published on 16 September, pointed to two subjects: computing and mathematics. School-aged pupils, the minister said, should learn the foundational knowledge that underpins AI literacy through those curricula. She also pointed to free National Centre for Computing Education courses to help teachers teach AI in computing.
It is a reasonable place to start. A pupil who understands data, patterns and probability is better equipped to ask why a system got something wrong. Computing can explain what a model is and how training data shapes what it produces.
Now ask where the other decisions live. Who teaches a student to check an AI-generated historical claim? Where do they discuss whether a generated image is fair evidence? Which department helps them decide when a chatbot's fluent explanation is replacing the thinking an assignment was meant to assess?
Those are not questions a computing department can answer alone.
The difference between a foundation and an entitlement
The wording of the parliamentary answer is careful. It talks about foundational knowledge that underpins AI literacy. It does not announce an AI literacy curriculum, a new statutory subject, or a timetable for teaching every aspect of responsible AI use. The minister said a new curriculum and assessment system is being developed and will be consulted on.
That leaves a practical gap for schools. One department may assume computing covers AI literacy. Computing may teach how systems work while never seeing the essay a pupil used one to write. English may deal with authorship, history with provenance and PSHE with privacy, yet nobody may know whether the pieces join up.
The result is familiar: a lesson here, an assembly there, a policy statement somewhere on the website. Pupils hear that AI can be wrong. They rarely have to show what they would do when it is.
Make the handoff visible
Before the curriculum consultation settles anything, a school can map three questions across existing subjects.
How does it work? Computing and maths can give pupils a basic account of training data, prediction, probability and limitations. The goal is enough understanding to challenge a model's authority.
How do I judge what it says? History can compare generated claims with sources. Science can test a plausible explanation against an experiment. English can examine voice, attribution and the point at which assistance becomes authorship.
When should I use it? Departments need their own answers. A tool that helps a pupil practise vocabulary may defeat the purpose of a task designed to assess independent writing. The acceptable boundary depends on what that particular lesson is trying to develop.
Put those answers in one curriculum map. Name the year group, lesson and evidence of learning for each. If the only evidence is that students were shown a slide, the map is not finished.
What to ask the consultation
The DfE has identified a foundation and a source of teacher training. The next questions are about sequence and responsibility: which knowledge must every pupil meet, at what age, and how will teachers outside computing be supported to apply it? How should schools show that pupils can evaluate AI in their own subjects rather than merely define it?
The parliamentary answer applies to England. Education policy in Scotland, Wales and Northern Ireland follows separate routes. For English schools, the useful move now is to let maths and computing provide the foundations, then make the rest of the curriculum's contribution explicit.
Source: UK Parliament, written answer on numeracy and AI skills, 16 September 2026.