What Oak Actually Published
On November 4, Oak National Academy, the English government-backed curriculum body, published a draft open knowledge graph modeling England's national curriculum. The graph represents subjects, topics, concepts and the relationships between them across 13 subjects and key stages 1 through 4, covering the bulk of primary and secondary education. It is released under the Open Government Licence v3.0 for the data and the MIT License for the underlying code, meaning any developer, commercial or otherwise, can use it without seeking Oak's permission or paying a fee.
Oak is explicit that this is a draft, not a finished production asset. Its structure and identifiers may still change, and the organization is actively inviting developers to test it and report what is missing before treating it as stable infrastructure. That caveat matters for procurement timing. Vendors integrating against this graph today should expect to maintain that integration as the model evolves, and institutional buyers should ask any vendor claiming to use it how it plans to handle version changes going forward.
The Problem: Every Vendor's AI Has a Different Idea of Aligned
Oak's CEO John Roberts framed the motivation in blunt terms: software can't reliably use the national curriculum as currently published, so each edtech company builds its own model, and those models don't always agree. That is a specific, practical complaint. When an AI tutoring tool or content recommender is asked a sequencing question, what a pupil needs to know before starting a new topic, it produces an answer that sounds authoritative regardless of whether it is actually correct, because it is working from whatever internal curriculum model its maker happened to build.
Roberts' own example makes the stakes concrete: pupils need to understand fractions before they can divide them. Get that prerequisite wrong inside an AI tool, and a student gets routed to content in the wrong order, with no obvious signal to a teacher that the sequencing was off. Mark Hodierne, Oak's senior machine learning engineer, put the underlying argument plainly: if AI is going to transform education, AI needs to know about education first. A shared, open model of what the curriculum actually requires is the fix Oak is proposing.
Why the Two-Layer Design Matters
The graph deliberately separates two things that most vendor-built curriculum models blend together. One layer holds the official national curriculum requirements, Oak's interpretation of the 2014 curriculum text. A second, separate layer holds Oak's own units, lessons and sequencing decisions, the specific teaching order Oak itself recommends. Roberts was careful to frame that second layer as one option among several: our sequencing is one evidence-informed route, not the only valid one, so it sits in its own layer.
That separation is the design choice that makes the graph usable by competitors rather than just by Oak's own products. A vendor can adopt the requirements layer as a shared baseline for what the curriculum demands, while building its own sequencing logic on top, without being forced to adopt Oak's specific pedagogical choices wholesale. For edtech companies wary of building on a competitor's proprietary teaching philosophy, that distinction is what makes the graph a standard worth adopting rather than a rival product to avoid.
A Government Signal, Not Just a Research Project
This release is not a standalone side project. Oak is simultaneously providing technical support to England's Department for Education as it develops a digital national curriculum intended to take effect from 2028, and Oak plans to fold that revised curriculum into the same graph once it exists. That gives the project a continuity most open edtech data releases lack: it is positioned to track the actual policy document England's schools will be required to follow, rather than a snapshot that goes stale once curriculum guidance changes.
For vendors selling into the England market specifically, that continuity is the real asset. A curriculum alignment claim built on a static, one-time export of curriculum data has a shelf life measured in policy cycles. A claim built on a graph that Oak is actively maintaining alongside the Department for Education's own 2028 curriculum work has a materially longer useful life, and vendors should factor that durability into which standard they choose to build against.
What This Means for Procurement Teams Evaluating Curriculum Claims
Every edtech vendor selling content-aligned or curriculum-mapped AI tools makes some version of the claim that its content matches what schools are actually required to teach. Until now, there was no independent way for a procurement team to check that claim beyond trusting the vendor's own documentation. An open, government-adjacent curriculum graph changes that. A procurement reviewer can now ask a specific, answerable question: does this vendor's alignment claim map to Oak's open requirements layer, and if not, why not.
That question should become standard in any RFP for curriculum-aligned AI tools serving English schools, and the underlying logic extends to other markets watching this release as a template. A vendor's inability to explain how its alignment model relates to a public standard is itself useful signal, pointing either to a proprietary model it cannot defend against scrutiny or to a product that was never rigorously mapped to the curriculum it claims to serve.
The Build vs Buy Calculus for Edtech Vendors
For vendors that have spent years and real engineering budget building and maintaining their own internal curriculum models, Oak's release is a direct challenge to that investment. Once a shared, actively maintained open standard exists, the case for continuing to fund a proprietary, hand-built model weakens considerably, particularly for smaller vendors who built that model mainly because no public alternative existed. Finance and engineering leaders who approved multi-year budgets for in-house curriculum mapping should revisit those line items now, before the next planning cycle locks in spending against a problem an open standard may soon handle at a fraction of the cost.
The practical move for most vendors is to adopt the requirements layer as a foundation and reserve proprietary investment for the sequencing and pedagogy layer, exactly the separation Oak's design encourages. That lets a vendor compete on teaching approach and product experience, the parts that actually differentiate it, while leaning on infrastructure the market as a whole now has reason to maintain collectively rather than duplicate thirteen times over. Vendors that make this shift early will free up engineering capacity for the product differentiation that actually wins deals, while slower movers keep funding maintenance work a government-backed standard is about to make redundant.
