The Oxford Rubric is a five-criterion framework for assessing whether the use of artificial intelligence in schools is safe, educationally sound, and ethically justified. Its five criteria — Safety, Efficacy, Accountability, Transparency, and Agency — give school leaders and teachers a structured way to evaluate any AI tool or practice before adoption.

AI is already in your school. Whether it arrived through a procurement decision, a teacher's curiosity, or a student's smartphone, generative AI tools are reshaping how secondary schools operate — faster than most policy frameworks can keep up with.

That gap between adoption and governance is exactly the problem the Oxford Rubric was designed to address. Published in January 2026 by Harwar, Keating, Valverde, and Burch, it offers schools something that's been notably absent from the AI-in-education conversation: a principled, practical checklist for deciding not just whether you can use AI, but whether you should.

This post unpacks what the Oxford Rubric actually says, how each of its five criteria applies in a secondary school context, and what your school can do right now to start using it.

Why Do Schools Need a Framework for AI in the First Place?

Secondary schools are under pressure from every direction. Vendors promise that their AI tools will save time, personalise learning, and reduce workload. Students are using AI tools regardless of school policy. And regulatory guidance — from Ofsted, the DfE, and beyond — is still catching up.

The result is a patchwork of ad hoc decisions made under time pressure, without a shared vocabulary or a consistent set of values to guide them.

The Oxford Rubric shifts the conversation from "Can we use this AI tool?" to "Should we use this AI tool, in this way, for this purpose?" That distinction matters enormously in a school setting, where the stakes include children's safeguarding, their development as independent thinkers, and the professional integrity of teachers.

What Are the Five Criteria of the Oxford Rubric?

The framework is intentionally technology-agnostic. It doesn't evaluate specific products; it evaluates practices. A use case should only be considered appropriate if it satisfies all five criteria. Failure on any single one signals that the use needs to be redesigned, restricted, or rejected.

1. Safety

Safety asks whether pupils are protected from physical, psychological, emotional, and informational harm when AI is used in any school scenario. It also covers how safeguarding duties are upheld and data protection obligations met.

In a secondary school, this means asking:

• Has a child-centred data protection impact assessment been completed?
• Are content filters enabled and appropriate for your students' age group?
• Is there adult supervision for any sensitive or high-risk interactions?
• Do clear protocols exist for when the AI fails, errors, or is misused?

A key principle here is that safety includes foreseeable misuse — not just intended functionality. An AI chatbot may be designed for learning support, but if a student can use it to access harmful content or if their personal data is processed without justification, the Safety criterion is not met.

2. Efficacy

Efficacy asks whether AI use demonstrably enhances teaching quality and learning outcomes — not just operational efficiency.

This is one of the most demanding criteria because it requires evidence. The rubric explicitly warns against adopting AI "primarily for efficiency or convenience." In a secondary school, that's a meaningful challenge. If an AI tool is generating student feedback faster, the Efficacy question is: is that feedback actually better? Is it leading to improved outcomes?

The rubric also raises the risk of over-scaffolding — where AI does so much for students that it reduces the cognitive effort and productive struggle through which real learning happens. For secondary teachers, this maps directly to long-standing debates about homework tools and the risk of replacing independent thinking with generated answers.

3. Accountability

Accountability addresses who is responsible for decisions, outcomes, and any harms that arise from AI use. The rubric is unambiguous: "AI systems do not bear moral or legal responsibility, but people and institutions do."

In practice, this means:

• School leaders must maintain ownership of AI-related decisions.
• Teachers must retain final professional judgement over all educational decisions.
• There must be a mechanism for reviewing, challenging, and overturning AI-influenced decisions.
• Governance cannot rely solely on vendor assurances.

The failure mode the rubric warns against is subtle but common: teachers feeling "compelled to follow AI recommendations without challenge," or accountability becoming "diffused or unclear in practice." If your school couldn't clearly answer who is responsible if an AI tool made a harmful decision about a student, your Accountability criterion isn't met.

4. Transparency

Transparency concerns the visibility and explainability of AI use to teachers, students, parents, governors, and inspectors.

The rubric's rationale here is grounded in trust: "When AI systems operate opaquely — making decisions, generating outputs or influencing outcomes without clear explanation — trust is eroded and informed consent becomes impossible."

For secondary schools, this has direct implications for how you communicate with parents and students about which tools are in use and why. It also means being honest about limitations. A writing feedback tool that sometimes generates inaccurate comments should be disclosed as such — not presented as an authoritative assessment.

5. Agency

Agency — perhaps the most distinctive of the five criteria — asks whether AI use preserves and strengthens the autonomy, judgement, and developing independence of both teachers and students.

The rubric frames this powerfully: "Education is fundamentally a human endeavour. AI should augment, not diminish, the professional agency of teachers or the developing independence of learners."

For secondary schools, this is particularly urgent. Adolescence is precisely the period when students should be developing independent thought, critical judgement, and intellectual confidence. An AI tool that makes it too easy to bypass that struggle — or that students treat as an authority rather than a tool — is actively working against the developmental purpose of secondary education.

The rubric also protects teachers. The risk of deskilling — where over-reliance on automation gradually erodes professional expertise — is real and underacknowledged in most AI adoption conversations.

How Does the Oxford Rubric Work in Practice?

The Oxford Rubric includes a simplified evaluation tool that rates each criterion as Green (appropriate), Amber (caution required), or Red (inappropriate). A use case is only acceptable when all five criteria are rated Green.

This traffic-light model makes it immediately practical for procurement decisions, lesson design, policy review, and Ofsted readiness. You can apply it to a specific tool ("Should we license this AI marking assistant?"), a specific practice ("Should Year 10 students be allowed to use AI for first drafts?"), or a whole-school policy.

Crucially, the framework supports context-sensitive judgement. A use case might be appropriate for Year 12 students working on extended essays and inappropriate for Year 7 students still developing foundational writing skills. The rubric doesn't give you a binary yes/no for AI in general — it gives you a principled way to evaluate each specific use in its specific context.

What Can Secondary Schools Do Right Now?

You don't need to wait for full implementation to start using the Oxford Rubric. Here are three practical starting points:

• Audit your current AI use against the five criteria. Pick one AI tool your school is already using and honestly assess it against Safety, Efficacy, Accountability, Transparency, and Agency. Where are the amber or red ratings? What would need to change to bring them to green?
• Use it before your next procurement decision. Before your school signs up to any new AI product, run it through the rubric. Ask vendors directly: How does your tool handle student data? What evidence of learning impact do you have? Can teachers opt out of AI recommendations?
• Build it into your AI policy. If your school is developing or reviewing an AI use policy, the Oxford Rubric's five criteria provide a ready-made governance structure. They also give you a language that staff, governors, and parents can engage with.

Frequently Asked Questions

Who created the Oxford Rubric?

The Oxford Rubric was created by Harwar, Keating, Valverde, and Burch and published in January 2026. It is being operationalised in partnership with 4i Studio (4i.studio) and is available at oxfordrubric.org.

Is the Oxford Rubric specific to UK schools?

The Oxford Rubric was developed with UK school structures and regulatory context in mind — including Ofsted alignment and statutory safeguarding duties — but its five criteria are applicable to any school system where children's welfare and educational integrity are priorities.

Does the Oxford Rubric apply to all AI tools, or only specific types?

The Oxford Rubric is intentionally technology-agnostic. It applies across classroom AI tools, assessment and feedback systems, behaviour and pastoral AI, administrative automation, and leadership decision-support tools. Any AI use in a school setting can and should be assessed against its five criteria.

What happens if an AI tool fails one of the five criteria?

Failure on any single criterion indicates that the use needs to be redesigned, restricted, or rejected. The rubric is not designed to produce partial approvals — all five criteria must be Green for a use case to be considered appropriate. An Amber rating signals caution and a need to mitigate risks before proceeding; a Red rating means the use is not currently appropriate.

The Bottom Line

AI adoption in secondary schools isn't slowing down. The question is whether it happens with intention, governance, and a clear ethical framework — or reactively, inconsistently, and at risk to the students and teachers involved.

The Oxford Rubric doesn't tell you which AI tools to use. It gives you the framework to make that judgement responsibly. In a landscape where vendors are eager to sell, regulators are still catching up, and the stakes involve children's development and safeguarding, that kind of principled structure is exactly what secondary schools need.

---

At thesecondary.ai, we help secondary schools navigate AI adoption with confidence. If you're developing an AI policy or evaluating tools for your school, explore our resources or get in touch.