The short answer
A school needs five things: a written position on AI (a policy or a policy section), evidence that staff have been trained, a decided and communicated stance on pupil use, confirmation from your filtering and monitoring provider on how it handles AI, and normal data-protection discipline applied to AI. The DfE's free training modules are the place to start for AI literacy. This guide explains what they cover, what they do not, and how to close the gap. The practical half they leave out, the classroom workflows, is what our Claude for Teachers course covers, for £99 per teacher with completion records for your training evidence.
Most designated safeguarding leads are not short of willingness on AI. They are short of a clear, evidenced answer to a simple governor question: what has the school actually done about it? "Staff know to be careful" is not an answer that survives contact with an inspection or a data-protection incident. The good news is that the requirements are more concrete and more achievable than the noise suggests, and a large part of the work is already provided free. The rest is a short, deliberate set of decisions written down.
What the DfE provides, and it is genuinely useful
In 2025 the Department for Education released free support for schools and colleges, developed with Chiltern Learning Trust and the Chartered College of Teaching. It is modular CPD covering what generative AI is, how to interact with it, how to develop safe use, and practical classroom applications, with slides, videos, planning templates and multiple-choice assessments. Completion can earn credit toward Chartered Teacher status, and the safe-use module is recommended for all staff. The Chartered College hosts the descriptor and access. If your school has done nothing formal yet, start there this term. It is free and official.
Alongside the training sit two documents worth knowing by name. The DfE policy paper Generative AI in education sets the department's overall position, and Generative AI: product safety expectations sets the safety benchmarks that pupil-facing AI products should meet, with a checklist schools can use to evaluate a tool before buying it. Together they answer the "is this product safe to put in front of children" question that procurement keeps asking.
Want to go deeper? Our Claude for Teachers course turns these tactics into a working skill with templates and exercises, for £99 per seat.
What KCSIE changed
Keeping Children Safe in Education brought generative AI clearly within the online-safety framework. The 2025 edition, statutory from 1 September 2025, strengthened filtering and monitoring expectations, added a self-assessment so schools can evidence an annual review against the filtering and monitoring standards, signposted the product safety expectations, and updated online-safety language to include misinformation and disinformation as harms. The 2026 edition, published in July 2026, comes into force on 1 September 2026 and continues in the same direction.
The practical implication for a DSL is the shift from intention to evidence. AI-generated harms such as deepfakes and AI-assisted bullying sit within existing child-protection procedures, filtering and monitoring arrangements are expected to address AI tools and the AI features now embedded in search and office suites, and the training expectation extends to staff understanding of AI risks. You need to be able to show that the school has assessed AI, trained staff, and decided its rules. Cite the gov.uk source directly rather than any paragraph number, because wording is refreshed each year.
The five-step compliance checklist
1. Write the position down. A standalone AI policy, or a clearly worded section in your online-safety and acceptable-use policies: approved tools, staff rules, pupil rules by key stage, and data red lines. Take it to governors so there is a dated record of approval.
2. Train staff and record it. The duty is evidenced by records: who was trained, on what, and when. The DfE modules cover awareness; hands-on workflow training covers the part staff actually ask about. Keep completion records either way.
3. Decide pupil use. Per key stage, reflected in your homework and assessment policy, and communicated to parents. Respect each tool's own minimum-age terms.
4. Ask your filtering provider one question in writing: how do you classify and handle generative AI sites and embedded AI features for each of our user groups? File the answer with your annual filtering and monitoring review.
5. Apply normal data protection. No pupil-identifying data in consumer AI tools, ever; education or enterprise tiers with a data processing agreement for anything deeper; and a data protection impact assessment before any pupil-facing AI product, checked against the DfE's product safety expectations. The ICO expects the risk assessment before deployment, not after an incident.
The three mistakes DSLs make
The first is treating a one-off staff briefing as done. Awareness that is not recorded, dated and refreshed does not evidence a training duty; keep a register and revisit it each year alongside your safeguarding updates. The second is banning AI outright, which simply pushes staff onto personal accounts on their phones, where the school has no controls at all. A clear approved-tools list with sensible rules is safer than a prohibition nobody follows. The third is confusing an AI policy with AI competence: a well-written policy that no member of staff knows how to act on protects the school on paper and helps no one in practice.
The pattern across all three is the same. Compliance is not a document you file; it is a set of decisions people understand and apply. That is why training and policy work best together, and why the strongest position for a school is a written stance, trained staff who can actually use the tools safely, and records that show both.
Where the free provision stops
The DfE modules answer "what is AI and why be careful." They deliberately do not teach the Monday-morning part: the differentiation, marking support and report-writing workflows that save staff time inside the safeguarding rules. That practical half is what our Claude for Teachers course covers, safeguarding-first, UK-specific, at £99 per teacher with per-staff completion records you can add to your training evidence. It is positioned deliberately as the step after the DfE modules, not a replacement for them.
Schools and trusts that want a structured rollout rather than scattered experiments sometimes go further than training. Prime AI Solutions runs AI opportunity assessments that map where AI genuinely helps and where it adds risk, and consulting to design the policy, the tool choices and the staff training as one coherent programme. The principle is the same whether you are a single school or a multi-academy trust: decide deliberately, write it down, train people, and keep the records.
Why Claude for Teachers is the best choice for teachers, in the UK and beyond
It is built by Prime AI Solutions, an AI consultancy led by Umar Din FCCA that trains professional teams across UK and MENA organisations. The same prompting method and tool-selection discipline we use in paid engagements, rebuilt for the classroom and wrapped in the safeguarding rules from the first lesson, not bolted on at the end.
- Personalised to the way teachers actually work: planning, differentiation, marking and reports
- Self-paced online, no coding or technical skills required
- Completion records for every teacher to evidence training, and KCSIE-ready for UK schools
- A free preview so you can judge the fit before you pay
- £99 per seat, cheaper than a single month of most AI subscription tools
We built Claude for Teachers because teachers everywhere deserve practical, safe AI training that hands their evenings back, not another compliance headache. It is designed around UK schools and their safeguarding duties, but the workflows work for teachers on any curriculum, part of a wider mission to put trustworthy, safeguarding-first AI skills into every staffroom, so the technology reaches classrooms on teachers' terms.
Explore Claude for TeachersClaude for Teachers: Plan, differentiate, mark and report, safely
The UK-schools course: lesson planning, differentiation, feedback and report scaffolds, with the safeguarding rules built into every lesson rather than bolted on. Self-paced, no coding required.