Irish Schools and the EU AI Act: Where to Begin

Anyone reading the EU AI Act for the first time would be forgiven for feeling out of their depth.
The good news is that most Irish schools can reach a defensible position before the August 2026 deadline with three pieces of evidence: an AI inventory built from a short staff survey, an Acceptable Use Policy updated for the Act, and a recorded staff literacy session under Article 4. The work fits inside a term and does not require legal training or a separate AI policy.
What follows walks through each step in turn; what it involves, who does it, and what "done" looks like. The wider picture is covered in a separate article. The Department of Education's AI in Schools External Advisory Taskforce, announced in April 2026,[1] will publish guidance in due course; until it arrives, this is where to start.
These steps move a school from starting out to being on track. Full compliance is a longer journey; this is where to start.
Step 1: What AI is actually being used?
This is the first question the EU AI Act implicitly asks of any AI system deployer (schools are deployers if staff use AI for work)[2]: What AI are staff already using for school work? Many schools won’t know right now. Not because anyone is hiding anything, but because AI arrived quickly and quietly, often embedded in tools staff already use. Only now has anyone been asked to document it.
A short staff survey is the practical answer. It doesn’t need to be extensive to be useful. A quick online form, sent to all teaching and non-teaching staff, asking about the tools they use and how they use them will provide the information needed. The goal is to have an overview of what’s actually happening, which is the starting point for every other decision.
The questions matter more than the format. A useful survey asks about specific tool categories: AI writing assistants, image generators, planning tools, AI features inside Microsoft 365 or Google Workspace, rather than asking “do you use AI?” in the abstract. Most teachers who use Copilot to draft a report don’t think of that as using AI. The questions have to be framed with this in mind.
Question 4 from the Leading Lights AI inventory survey (Microsoft 365 version). Highlights are editorial annotations; they do not appear in the version sent to staff.
Send the survey with a short covering note. Keep the tone curious rather than investigatory; the aim is to get an accurate picture, not catch anyone out. A week should be sufficient time to gather responses. The results become your school’s AI inventory: a simple record of what’s in use, by whom, and in what context.
The aim is to get an accurate picture, not catch anyone out.
It's possible some of what you find may require attention under Article 5 of the Act, the prohibited practices that have been in effect since February 2025.[3] We’ll cover the shadow AI question in more depth, including the tools most likely to surface in your results and recommended reactions, in a future article in this series.
Leading Lights has a ready-built version of this survey, along with a processing spreadsheet that turns the results into a readable inventory. Get in touch if you’d like a copy.
You now know what's in use. That's the foundation everything else builds on.
Step 2: Review your Acceptable Use Policy
Your school’s Acceptable Use Policy is the document that governs how digital technology is used on school premises and networks. It’s also, under the EU AI Act, the natural home for AI governance.
The Department of Education has been clear on this: a separate AI policy is not required. Schools should amend existing policies rather than creating new ones.[4] The Webwise AUP Generator, updated in 2025, includes 21 AI-specific statements covering ethical use, data handling, and student use options.[5] If your AUP hasn’t incorporated any of those statements, that’s a good place to start.
Under the EU AI Act, a school's Acceptable Use Policy is the natural home for AI governance.
The second thing is less obvious. Webwise addresses transparency and human oversight, but in general terms rather than specific to deployers under the Act. The gap is in the detail: high-risk AI classifications, formal human oversight assignments with documentation, and AI Act-specific notifications when students interact with certain systems. Webwise gives schools a good foundation but doesn’t yet provide full coverage.
This is not meant as a criticism of the Webwise resource. It was written before the Act’s obligations were fully clear. Starting with what Webwise provides then adding the specific AI Act layer on top gets a school into a good position.
| Webwise statement (as written) | Strengthened for EU AI Act compliance |
|---|---|
| “The school is committed to ensuring that all AI tools used in our school comply with relevant data protection regulations, including GDPR.” | “The school is committed to ensuring that all AI tools used in our school comply with relevant data protection regulations, including GDPR, and with the obligations placed on deployers under the EU AI Act 2024. Where AI systems interact with students or parents, the school will ensure those interactions are disclosed. Where AI systems are used to support decisions about students’ learning, assessments, or progression, a member of staff with appropriate competence will retain oversight of those decisions and will be capable of identifying, assessing, and, where necessary, overriding them.” |
The strengthened version adds what the Webwise statement leaves out: the Act by name, disclosure of AI interactions, and the human oversight requirement.
Once the AUP is updated, it needs to go to the Board of Management for ratification, after which it becomes school policy rather than a working document. The AUP will be covered in full in a later article, including the specific AI Act obligations that need to be reflected and how to frame the update for a Board meeting.
Step 3: Book a staff literacy session
Article 4 of the EU AI Act has been in effect since 2 February 2025. It requires deployers, such as schools, to take appropriate measures to ensure their staff have sufficient AI literacy to understand the tools they are using and the risks those tools carry.[6] That is a legal obligation, not a professional development recommendation.
The practical question is what “sufficient literacy” looks like and how to document it. The answer doesn’t have to be complicated. A single staff session, properly structured and recorded, is a reasonable starting point for most schools. It won’t satisfy the full obligation indefinitely because AI literacy will be an ongoing requirement as tools and practices evolve. It does however demonstrate that deliberate steps have been taken, which is what the Act is asking for.
The session should cover what AI tools are in use in the school (using the inventory from Step 1 as the basis); what the rules are around staff use, student use, and data (using the updated AUP as the reference); and what specific practices are prohibited under Article 5, explained in plain terms. Sixty minutes is enough if the session is well prepared.
St. Brendan’s NS
Staffroom · 4.00–5.00pm
- Survey results: tools currently in use across the school
- Updated AUP: what’s allowed, what needs authorisation, what’s not permitted
- Article 5 prohibited practices: emotion recognition, manipulation of minors, social scoring: definitions and examples
A brief written record stating the date, who attended, what was covered, and the AI tools discussed is your Article 4 evidence if it is ever requested. Keep it in the same folder as your updated AUP and your AI inventory.
Your school has taken documented, deliberate steps. That is what the Act requires at this stage.
What this doesn’t cover
An inventory, an updated AUP, and a staff literacy session move a school into a defensible position. They do not address every obligation under the Act.
High-risk AI systems, which include tools used for grading, learning pathway decisions, admissions, or exam proctoring,[7] carry additional requirements: formal human oversight, records kept for at least six months, and in some cases a Fundamental Rights Impact Assessment (FRIA) before deployment.[8] Schools that use tools in any of those categories would have them revealed by the survey. Articles to follow will cover those obligations in detail.
The Board of Management’s specific governance responsibilities, and what they need to formally approve and record, will be covered in a later article too.
None of that needs to happen first.
What you’ll have by the end
A survey can be filled in 10 minutes; evaluating the results is done within a week. An AUP review takes a meeting. A literacy session takes an afternoon. Between them, they produce the documented evidence the Act is looking for: an inventory, a ratified policy, and an Article 4 record.
A survey can be filled in 10 minutes; evaluating the results is done within a week. An AUP review takes a meeting. A literacy session takes an afternoon.
The Act reads like deep water at first, but navigating it is just about taking that first stroke.
Common questions
Do schools have to complete all three steps at once?
No. The three steps are a sequence, not a single event. The survey comes first because it informs the other two. The Acceptable Use Policy review and the staff literacy session can follow across the same term, as time allows. What matters is that each step is recorded once it is done.[9]
Is ChatGPT allowed in schools in Ireland?
There is no blanket ban. The school's Acceptable Use Policy decides what is and isn't permitted, and use must comply with the EU AI Act and GDPR. Staff and student use are treated differently. Personal data should not be entered into public tools.
Does a school need a separate AI policy under the EU AI Act?
No. The Department of Education has confirmed schools should update existing policies rather than draft a standalone AI policy.[4] The Acceptable Use Policy is the natural home; the school's data protection policy, code of behaviour, and Bí Cineálta procedures may also need light revision.
What is a high-risk AI system in a school context?
Tools used for grading, learning pathway decisions, admissions, or exam proctoring fall under Annex III of the Act and are classed as high-risk.[7] Schools using these face additional obligations, including formal human oversight and, in some cases, a Fundamental Rights Impact Assessment. The pillar article covers high-risk classification in more detail, and a dedicated article will follow. Schools using these face additional obligations: formal human oversight, six-month record retention, and in some cases a Fundamental Rights Impact Assessment (FRIA).[8]
What does an AI literacy session need to cover?
For most schools a single recorded session is a reasonable starting point. It should cover the AI tools currently in use, drawn from the inventory; the rules on staff use, student use, and data, drawn from the updated AUP; and the practices prohibited under Article 5, explained in plain terms. Keeping a short record of the session, with date, attendees, and topics, is what demonstrates the Article 4 obligation has been met.[6]
Who regulates the EU AI Act in Ireland?
Ireland uses a distributed model under S.I. No. 366/2025, with multiple regulators coordinated by the AI Office of Ireland. For schools, the Data Protection Commission is the most likely relevant authority because most AI issues in schools touch personal data.[10]



