# Does a School Need a DPIA for Every AI Tool? A Screen for the DPO

Canonical URL: https://blog.theaieducator.io/posts/dpia-for-ai-tools-in-schools
Publication: Dan Fitzpatrick Insights
Author: Dan Fitzpatrick
Topic: AI Governance and Responsible Leadership
Published: 2026-11-09T07:20:00.000Z
Modified: 2026-11-09T07:12:09.789Z

Not every AI tool needs a full DPIA, but every one needs a recorded decision. A five-question screen helps a school's data protection lead decide in minutes, and shows why the supplier's own assessment never covers your use.

## In brief

A school does not need a full DPIA for every AI tool, but it needs a recorded screening decision for each one, because AI counts as innovative technology and children's data raises the risk. The supplier's DPIA assesses the product, not your use. Five questions let a data protection lead decide in minutes which tools go to the DPO.

## Key takeaways

- Every AI tool a school uses needs a recorded screening decision, even when a full DPIA turns out not to be required.
- The ICO lists AI as innovative technology, which requires a DPIA when combined with another criterion such as children's data.
- A supplier's DPIA assesses the product; only the school can assess what its own staff and students put in and what the output decides.
- The DfE's product safety standards, first published January 19, 2026, place the DPIA duty on developers and do not tell schools to skip their own.
- The Five-Question Screen, Dan's suggested way of thinking, asks about personal data, supplier retention or training, decisions about named people, monitoring, and change.
- Any yes on the screen sends the tool to the data protection officer before go-live; five noes earn a one-line record with an owner and a review date.
- The ICO's reprimand of Chelmer Valley High School shows the regulator expects a DPIA before processing starts, not afterwards.

Not every AI tool a school uses needs a full data protection impact assessment (DPIA), but every one needs a recorded decision about whether it does, with a name beside it. Where children's personal data goes into a tool, one law firm that advises English schools says an assessment is almost certainly required. And the supplier's own assessment does not stand in for yours.

Picture a shared folder called "AI tools: data protection." It holds eleven PDFs, and every one was written by a supplier. Nobody at the school has written a page of their own. If an inspector, a parent or the Information Commissioner's Office (ICO) asked what the school had weighed before a tool went live, the folder would answer for the suppliers and say nothing about the school. That folder is hypothetical, but I would be surprised if you have never seen its cousin.

This article is about England and the UK GDPR, because that is where the DPIA duty sits. If you lead a US district, the same question arrives under FERPA and state student-privacy laws, which work differently and which I do not cover here. I am not a lawyer. Your data protection officer (DPO) makes the legal call. What follows is how I would help a DPO make it quickly.

## Does every AI tool need a DPIA?

No, but every AI tool needs a screening decision on record, because the law asks for an assessment wherever processing is likely to put people at high risk, and AI sits close to the line.

A DPIA is the written assessment UK GDPR requires before processing that is likely to result in a high risk to individuals. The Department for Education (DfE) cites Article 35 of the UK GDPR for this in its [Generative AI: product safety standards](https://www.gov.uk/government/publications/generative-ai-product-safety-standards/generative-ai-product-safety-standards), first published January 19, 2026. The ICO's own [list of processing likely to result in high risk](https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/accountability-and-governance/data-protection-impact-assessments-dpias/examples-of-processing-likely-to-result-in-high-risk) names AI and machine learning as innovative technology. It says a DPIA is required for innovative technology when it is combined with another listed criterion. Children's data and automated decision-making appear among those criteria.

A law firm's advice lands in the same place. Browne Jacobson, in its March 4, 2025 guidance [AI adoption in schools: meeting ICO expectations](https://www.brownejacobson.com/insights/ai-adoption-in-schools-meeting-ico-expectations), says all projects should be screened to see whether a DPIA is required. It adds that where children's personal data is processed, that is "almost certainly a high-risk processing activity." It also says Ofsted will want to see evidence of the assessment. That last point is the firm's view, and I would treat it as a sensible expectation, not a stated inspection rule.

What skipping the step looks like is on the record. In its 2024 [reprimand of Chelmer Valley High School](https://cy.ico.org.uk/media/action-weve-taken/reprimands/4030350/20240702-cvhs-reprimand.pdf) in Essex, the ICO found the school had introduced facial recognition for its cashless catering system in March 2023 without a DPIA. The ICO's finding was that the school "has therefore failed to complete a DPIA where they were legally required to do so." The school completed one in November 2023, and the ICO welcomed that. This was biometric technology, not a chatbot, so it is not a precedent for AI tools. It does show what the regulator expects to see, and when: before the processing starts.

## Why does the supplier's DPIA not cover your school?

Because a supplier assesses its product, and the risk in a school sits in how the product is used.

The DfE standards are aimed mainly at suppliers. They say developers should conduct a DPIA during development and across the life of the tool. The page does not tell schools to run their own, which is a gap in what it says, not a statement that schools are excused. Under the same legislation a school that decides why and how pupil data is used carries its own responsibilities. Browne Jacobson makes the related point that a school cannot hand responsibility for AI decision-making to the vendor.

My reading is this. A supplier can describe what its software does with the data it receives. It cannot know what your teachers will paste in, which year groups will use it, or whether its output will shape a grade or a support plan. Take one tool and two uses (both hypothetical). A teacher uploads her own slides and asks for a ten-question quiz. Another teacher uploads Year 9 essays with names attached and asks for target grades. The software is identical and the data protection position is not. Only the school can see that difference, so only the school can assess it.

## What is the five-question screen?

It is a way for a busy data protection lead to decide in five minutes whether a tool needs a full DPIA, and I offer it as my suggested way of thinking, not as the legal test.

I call it the Five-Question Screen. Ask five questions of any AI tool and the use proposed for it. A yes to any one means the DPO sees the tool before it goes live. Five noes mean a one-line record of what the tool is for, who approved it and when it will be looked at again.

| Question | What it is really asking | Rough link to the ICO's triggers |
|---|---|---|
| 1. Could personal data go in, including free text that identifies a child? | Whether anyone's data enters the tool at all. Deleting a name does not remove a child from a sentence about a small class. | Children's data |
| 2. Could the supplier keep it or train on it? | Whether data leaves your control. The DfE standards say products should not use personal data for commercial purposes, which includes model training, without a confirmed lawful basis. | Not a listed trigger on its own; it decides who ends up controlling the data |
| 3. Does the output shape a decision about a named person? | Whether the tool is part of a grade, a support plan, a behavior record or a hiring choice. | Automated decision-making and profiling |
| 4. Does it record, watch or listen? | Whether transcripts, monitoring or proctoring create a record of people going about their day. | Monitoring |
| 5. Is anything new or changed since we last looked? | A new feature, model or set of terms, or a new group of users, such as students where only staff used it before. | Innovative technology |

The right-hand column is my shorthand for the kinds of triggers the ICO lists. It is not a substitute for reading the ICO's list or for your DPO's judgment. What the screen does is turn an abstract duty into five questions a head of department could answer standing in a corridor.

Run it on the two uses above. The slides-to-quiz request gets five noes: no personal data, nothing kept, no decision about a person, no recording, nothing changed. It earns a one-line record. The essay-feedback request answers yes to questions 1 and 3, because it carries names and feeds target grades, and probably to 2 as well until the supplier's terms are read. That tool goes to the DPO before anyone uses it with a class.

## What I tell leadership teams about the paperwork

The mistake I see most often is treating a supplier's documentation as the school's assessment, and the second most common is doing the screen once and never again.

Across the leadership teams I work with, the supplier's security summary gets filed and the file gets called the DPIA. The folder looks complete. Then a vendor adds a feature, changes its terms or starts a new product tier, and nothing in the school's records changes, because nothing in the school's records was ever the school's. Question 5 on the screen exists for this reason. The same logic runs through [your approved AI tool just changed](https://blog.theaieducator.io/posts/when-an-approved-ai-tool-changes), and it is why a tool should be re-screened whenever it is reapproved, for example at the term-end review described in [which AI tools a school should stop using](https://blog.theaieducator.io/posts/which-ai-tools-should-a-school-stop-using).

I have sat at both ends of the approval table, as a former assistant headteacher asking for paperwork and now as a school trustee reading it. A trustee does not want the supplier's brochure. A trustee wants to know who looked at this tool, what they decided and when they will look again. A screening record answers all three in a line.

## What does a DPIA for an AI tool need to contain?

A usable one is short, specific to your use, and signed by a named person, because its job is to show that someone thought before the data moved.

A DPIA describes the processing, tests whether it is necessary and proportionate, identifies the risks to the people involved and records how those risks will be reduced. For an AI tool in a school I would want these lines to be answerable in plain words:

- What the tool is for, in a sentence naming the job and the year groups or staff involved.
- What goes in, what comes out and whether either contains personal data.
- Whether the supplier is acting as a controller or a processor, and what its written terms say about retention and model training. Browne Jacobson advises schools to be clear on the controller question before they sign.
- What the risks to children are if it goes wrong, including bias or a wrong output being acted on.
- What the school will do to reduce them: which data is kept out, who reviews outputs before they affect a child, and what staff are told.
- What the DPO advised, who signed, and the date the assessment will be reviewed.

If the assessment runs to more than two pages for a single tool, it has probably stopped being used by the people it is meant to protect.

## What should a data protection lead do this half term?

Start with the tools already in use, not the ones in the pipeline, because the gap between approved and actual is where the unassessed processing lives.

1. List every AI tool staff or students touch, approved or not. The staff survey in [how should a principal survey staff about AI](https://blog.theaieducator.io/posts/staff-ai-survey-questions-school-leaders) is a quick way to find the ones that are not on the list.
2. Run the Five-Question Screen on each. Expect most to produce five noes. Write the one-line record for those and move on.
3. Send every yes to the DPO, student-facing and decision-shaping tools first. Agree who owns each assessment, so that the name on it is a person and not a committee.
4. Link the record to the approved list described in [what schools should require before approving an AI tool](https://blog.theaieducator.io/posts/approving-ai-tools-for-schools), so one list carries both decisions.
5. Put a review date on every record. A tool that changes gets re-screened, and a tool that nobody can account for comes off the list.

Keep the rule about personal data as plain as the screen. [Can staff put student data into ChatGPT](https://blog.theaieducator.io/posts/can-staff-put-student-data-into-chatgpt) sets out what to tell teachers, and [who is accountable when AI gets it wrong](https://blog.theaieducator.io/posts/who-is-accountable-when-ai-gets-it-wrong) covers what happens after something does.

If your leadership team is working through these questions, this is the kind of work I support through [AI strategy sessions for schools](https://theaieducator.io/ai-strategy-for-schools?utm_source=blog.theaieducator.io&utm_medium=referral&utm_campaign=dpia-for-ai-tools-in-schools) and board and leadership sessions through [keynotes and workshops](https://theaieducator.io/education-keynotes?utm_source=blog.theaieducator.io&utm_medium=referral&utm_campaign=dpia-for-ai-tools-in-schools). My newsletter, [on the AI Educator site](https://theaieducator.io/?utm_source=blog.theaieducator.io&utm_medium=referral&utm_campaign=dpia-for-ai-tools-in-schools#newsletter), covers the changes worth acting on as they arrive.

## Sources and further reading

- Department for Education, [Generative AI: product safety standards](https://www.gov.uk/government/publications/generative-ai-product-safety-standards/generative-ai-product-safety-standards), GOV.UK, first published January 19, 2026 (England).
- Information Commissioner's Office, [Examples of processing likely to result in high risk](https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/accountability-and-governance/data-protection-impact-assessments-dpias/examples-of-processing-likely-to-result-in-high-risk), ICO guidance (undated page).
- Information Commissioner's Office, [Reprimand to Chelmer Valley High School](https://cy.ico.org.uk/media/action-weve-taken/reprimands/4030350/20240702-cvhs-reprimand.pdf), 2024.
- Claire Archibald and Dai Durbridge, [AI adoption in schools: meeting ICO expectations](https://www.brownejacobson.com/insights/ai-adoption-in-schools-meeting-ico-expectations), Browne Jacobson, March 4, 2025.

*Dan Fitzpatrick is a keynote speaker, Forbes contributor and the founder of The AI Educator, who has trained more than 150,000 educators across 30+ countries. [About Dan](https://www.theaieducator.io/about).*

## Frequently asked questions

### Does a school need a DPIA for every AI tool?

No, but every AI tool needs a recorded screening decision. A full DPIA is required where processing is likely to be high risk. The ICO lists AI as innovative technology, which needs a DPIA when combined with another criterion, such as children's data. Your data protection officer makes the final call.

### Is the supplier's DPIA enough for a school?

No. A supplier assesses its product, and the DfE's product safety standards expect developers to run a DPIA across the tool's life. The school's risk depends on what staff and students put in and what the output decides, so the school needs its own record of that use.

### When must a school complete a DPIA before using AI?

Before the processing starts. The ICO's 2024 reprimand of Chelmer Valley High School found it had introduced facial recognition without a DPIA it was legally required to do. Browne Jacobson advises that where children's personal data is processed, the processing is almost certainly high risk.

### What should a school's AI tool screening check?

Five things: whether personal data goes in, whether the supplier could keep it or train on it, whether the output shapes a decision about a named person, whether the tool records or monitors people, and whether anything has changed since the last review. Any yes goes to the data protection officer.

### How often should a school review an AI tool's data protection record?

Whenever the tool changes and at a fixed date set in the record, such as the end of each term. A new feature, new terms or a new group of users can alter the risk, so every record should carry a named owner and a review date.

---
Source: [Does a School Need a DPIA for Every AI Tool? A Screen for the DPO](https://blog.theaieducator.io/posts/dpia-for-ai-tools-in-schools)
Publisher: [The AI Educator](https://theaieducator.io)
