AI Marking Software for School Data Protection Officers

When a department asks to bring in AI marking software, the request usually lands on the data protection officer's desk before it lands anywhere else. And rightly so: student work is personal data, AI processing is exactly the kind of activity that warrants careful thought, and a tool adopted without that scrutiny is a risk the whole school carries. If you are the DPO — or the business manager, deputy, or governor who holds that responsibility — this guide looks at AI marking software from a data protection standpoint and sets out how GradeOrbit is built to make your due diligence straightforward rather than a roadblock.

Start With the Right Question

The instinctive first question is "is it secure?" The more useful one is "what happens to a piece of student work from the moment a teacher uploads it to the moment the result comes back?" Data protection lives in that sequence: what is processed, where, by whom, for how long it is kept, and on what lawful basis. A vendor that can answer that sequence clearly is one you can assess; a vendor that answers in marketing language is one you cannot.

It also matters that the answer is the right shape for a school. Many general-purpose AI tools are built for consumers or businesses and bolt on an education story afterwards. A tool designed for UK schools should already be set up around the realities you work within — the need to return to a student's marked work later, retention windows that respect appeal and results-challenge timeframes, and a clear processor relationship with the school as controller.

Two Models of Handling Student Work

GradeOrbit is worth understanding because it deliberately offers two different data handling models for two different situations, and a DPO should know which one applies.

For individual teachers using a personal account, the model is never save. Student work is processed to produce the marking or detection result and then discarded — it is not written to any database or storage. Before uploading, the teacher redacts names and identifying details by drawing black boxes over them, and that redaction is burnt into the image in their browser, so the identifying information never leaves their device. For a teacher experimenting on their own, this is the lowest-footprint arrangement possible: nothing is retained, and the personal data is minimised before processing even begins.

For a school account, the model is the deliberate inverse, and this is the one most relevant to you. Under a school subscription, student work is persisted — because the institutional use case requires it. Teachers need to return to a student's marked work and results after the fact, results feed into the school's own analytics, and a moderator or appeal may need the evidence weeks later. That persistence is not an oversight; it is a designed feature, and it operates under a signed Data Processing Agreement between the school and GradeOrbit. The school is the controller; GradeOrbit is the processor acting on the school's documented instructions.

What the DPA and Retention Window Cover

The Data Processing Agreement is the document your assessment should centre on. It is what turns "we use an AI tool" into a defined controller–processor relationship with obligations on both sides. For the school account, the things a DPO will want to confirm are the ones the DPA exists to set out: that student work is processed only to deliver marking and AI detection, that it is held under defined retention rules rather than indefinitely, and that there is a clean exit when the school offboards.

Retention is windowed rather than open-ended. Persisted student work is kept against a school retention period aligned to UK appeal and results-challenge timeframes, with shorter per-run archiving so that older assessment runs have their stored imagery cleared automatically once they are no longer needed. When a school leaves, offboarding wipes the school's stored work and records. The principle a DPO should test for is the one that matters under UK GDPR's storage limitation: data is kept for as long as the educational purpose requires, and then it goes.

The Questions to Put to Any AI Marking Vendor

Whether you are assessing GradeOrbit or comparing it against alternatives, a consistent checklist keeps the conversation grounded. Ask where the processing happens and which sub-processors are involved, since AI marking typically relies on a large-language-model provider in the chain. Ask whether there is a signed DPA available for the institutional tier, and read what it actually commits to. Ask how long student work is retained and what triggers its deletion. Ask what happens to the data when the contract ends. And ask whether the tool stores real student identities or anonymises — and if it stores them, on what basis and under what safeguards.

A vendor built for schools will have crisp answers to all of these because the questions are not an afterthought for them either. The aim of the exercise is not to find a tool with zero data footprint — a marking tool that returns to a student's work necessarily keeps something — but to find one whose footprint is defined, documented, time-limited, and contractually bound.

Helping Your Teachers Get It Right

Part of the DPO role is making the compliant path the easy one. With GradeOrbit, the solo and school models do that work for you: a teacher trialling the tool privately is on the never-save, redact-first model by default, and a school rollout runs under the DPA with persistence handled inside the agreed retention rules. Your guidance to staff can be simple — personal experimentation stays on personal accounts and never carries un-redacted identifying data; anything that needs to persist student work runs through the school account under the DPA. That clarity is easier to train and easier to audit than a single ambiguous arrangement.

It also dovetails with the wider integrity picture. Many schools adopt the same tool for AI detection as for marking, and a DPO assessing one is usually assessing both; our guide to whether AI detection tools are safe to use in schools covers that side, and the SLT questions guide frames the procurement conversation for the leadership team you will be advising.

Try GradeOrbit With Your Due Diligence In Hand

If your school is weighing AI marking and you want to assess it properly before any rollout, GradeOrbit is built to make that assessment straightforward — with a clear two-model approach to student work, a Data Processing Agreement for the institutional tier, and retention that respects UK timeframes rather than keeping data indefinitely. A department can trial the marking workflow on a real class set with the free credits on a new account, with no commitment, while you run your checks.

Visit gradeorbit.co.uk to learn more and get started, and bring your data protection questions — they are exactly the ones the school model was designed to answer.