AI in Curriculum: What DfE Review Means for Schools

A pupil can now use generative AI to plan an essay, write code, revise a topic or create an image within seconds. Teachers and school leaders are also testing it for lesson planning, feedback, administration and tailored support. The AI in the curriculum DfE review matters because it could shape what pupils learn, how staff work, how schools manage risk and which skills young people need after school.

The central question is clear: should schools only control pupils’ use of AI, or teach them to understand, test and use it responsibly? The answer is starting to emerge, but schools must separate current duties from recommendations and future reforms.

Why the DfE review could reshape AI education

What the review is examining

This is part of the wider Curriculum and Assessment Review for England, chaired by Professor Becky Francis and commissioned in July 2024. Its final report on curriculum and assessment was published on 5 November 2025 after evidence from teachers, pupils, families, employers and subject bodies.

The review covers education from ages 5 to 19, including national curriculum subjects, qualifications and assessment. It is not a technical review of generative AI alone. Instead, it considers how schools should respond to wider changes in digital technology, media, work and society.

The panel recommends clearer AI content within Computing, stronger digital and media literacy, and assessment that protects fairness. These are recommendations, not all current statutory requirements.

Why schools cannot treat AI as temporary

AI already affects homework, revision, research, writing, coding and careers. The DfE says teacher-facing uses offer more immediate benefits, while evidence about pupil-facing use remains limited. Its guidance on generative AI in education warns that outputs may be false, biased, unsafe, out of date or taken from copyrighted material.

A school policy cannot remove these issues by ignoring them. Pupils need enough subject knowledge to spot weak answers, check evidence and understand when human judgement must take priority.

What AI in the curriculum could change

Computing could move beyond coding

The current Computing curriculum already covers algorithms, programming, data, networks, digital content and online safety. It contains limited explicit reference to AI. The review recommends greater clarity at each key stage and direct coverage of AI within a refreshed Computing curriculum.

Possible content includes training data, machine learning, model limits, bias, automation, cybersecurity and the environmental cost of computing. The Government has confirmed plans for a broader Computing GCSE to replace the narrowly focused Computer Science GCSE, while it explores a new 16-18 qualification in data science and AI.

Schools should not treat these future qualifications as current requirements. They should compare any local changes with the existing statutory programme of study.

English, humanities, science and careers will also be affected

An AI-generated historical claim can become a useful lesson in checking sources, context and bias. In Science, pupils can compare an AI explanation of photosynthesis with a trusted textbook, then correct its errors. English lessons can examine authorship, persuasive language, copyright and the difference between editing and replacing a pupil’s own thinking.

The review also links AI with stronger digital and media literacy. Careers education should build skills that last beyond individual tools: asking precise questions, checking results, solving problems, communicating clearly and making ethical decisions.

The practical implications for teachers and leaders

Staff development and acceptable-use policies

Teachers need time to learn how generative AI works, recognise hallucinations, protect confidential information and check generated content. Training should also cover prompt design, bias, age limits and the safe use of pupil work. The Government’s response points to further support through the National Centre for Computing Education.

School policies should define permitted and prohibited uses rather than rely on a blanket ban. They should cover disclosure of AI assistance, data protection, safeguarding, plagiarism, assessment rules, consequences for misuse and links with behaviour and online-safety policies.

The DfE says schools may choose different models, including staff-only use or supervised pupil use in selected year groups. Ofsted does not inspect a tool in isolation. It may consider how leaders assessed risks and the effect on pupils.

Assessment should reveal thinking

Unsupervised essays and homework are harder to authenticate when pupils can generate polished text. The answer is not to discard every form of coursework. It is to gather better evidence of learning through drafts, annotated sources, oral questioning, in-class writing, process portfolios and short reflections on tool use.

The Curriculum and Assessment Review says exams should remain central because they reduce malpractice risks. It also recommends that DfE and Ofqual continue examining assessment methods that protect validity while managing generative AI.

How to teach pupils to use AI safely

Primary foundations and secondary depth

Primary pupils need age-appropriate teaching about privacy, reliable information and the fact that computers do not understand the world as people do. They should know to ask a trusted adult before using unfamiliar tools and should not enter personal details into public services.

Secondary pupils can study prompts, hallucinations, deepfakes, copyright, personal data and academic honesty. A strong activity is to give pupils a flawed AI answer, ask them to identify unsupported claims and improve it using authoritative sources.

The Government’s 2026 Media Literacy Action Plan supports stronger teaching about algorithms, misinformation and AI. Updated RSHE guidance, used from September 2026, also includes AI, deepfakes and social media.

SEND provision must protect agency

AI may support access through speech-to-text, translation, text simplification and alternative ways to express ideas. Yet inaccurate adaptations, privacy risks and over-reliance may reduce opportunities to build core skills.

Schools should assess each pupil’s needs and connect AI support with existing reasonable adjustments. The tool should increase access without removing choice, effort or ownership from the learner.

Safeguards and evidence before adoption

Data, accuracy and procurement

Leaders should involve the school’s data protection officer before staff or pupils use a tool. Check account requirements, age limits, third-party processing, retention, image uploads and the contents of prompts. The DfE advises against entering personal data into generative AI tools, and schools should also consult ICO education guidance and their privacy notices.

Test tools with real curriculum tasks before wider use. Check for factual errors, cultural bias, discriminatory assumptions, unsafe content and inaccessible language. Educational branding does not remove the need for human review.

Procurement should start with the learning goal, not the novelty of the product. Consider evidence of impact, cost, staff workload, security, accessibility, interoperability, vendor transparency and exit arrangements. A small pilot with clear success measures is safer than rapid school-wide adoption.

How to build an effective AI curriculum plan

Map provision and begin with low-risk tasks

Audit Computing, English, Science, Citizenship, RSHE, PSHE, careers, safeguarding and assessment. Create a progression showing what pupils should know, understand and do at each phase.

Schools can begin without sending pupil data to public tools. Teachers might present an AI-generated answer for pupils to fact-check, compare human and machine explanations, or ask classes to repair an inaccurate response using trusted sources.

Involve the whole school community

Staff, pupils, families and governors can identify practical concerns that a policy team may miss. Families should know which tools are approved, what homework rules apply, when pupils must disclose AI assistance and how privacy is protected at home.

Track pupil understanding, staff workload, assessment concerns, safeguarding incidents and accessibility outcomes. Review the plan when the DfE publishes new curriculum documents, consultation results or assessment rules.

What schools should watch as policy develops

Statutory duties and future changes

The review’s recommendations do not automatically change the law or current national curriculum. Schools should distinguish legislation, statutory guidance, non-statutory guidance, consultation proposals and commentary. Each DfE source should be checked for its status and publication date.

The Government has confirmed that the refreshed national curriculum will be digital and machine-readable, with full implementation planned for first teaching in September 2028. A consultation on revised programmes of study is expected later in 2026, with publication planned for spring 2027.

A gradual transition is more likely

Schools do not need to wait before teaching sound AI literacy. They should, however, keep changes consistent with current statutory duties and official guidance. Watch DfE publications, Ofqual assessment updates, Ofsted communications, ICO advice and recognised subject associations.

Conclusion

The DfE review points towards AI literacy as part of a wider, knowledge-rich curriculum. Computing is likely to gain clearer AI content, while English, Science, History, Citizenship, RSHE and careers education will help pupils test information and act responsibly.

The immediate priorities are practical: audit current provision, train staff, update acceptable-use rules, protect personal data, strengthen assessment and trial low-risk activities. Schools that prepare well will not be those using the most AI tools. They will be those teaching pupils to question outputs, verify evidence, make ethical choices and remain accountable for their work.