UK AI in Education and Training: Policy Tracker
A source-led tracker for current UK policy on AI in education and training — covering DfE, Ofqual, Ofsted, ICO, Skills England, JCQ and wider UK regulation. Each entry distinguishes an enforceable rule from guidance, research or a future-effective inspection update. Last reviewed: 15 July 2026.
How to use this tracker
This tracker summarises the current policy position of each body. It is a starting point for understanding the landscape — always check primary sources before making compliance decisions, as guidance evolves. Links to primary sources are provided in Section 2.
Section 1: Policy Status Summary
Current position of key bodies on AI in education and training as at 15 July 2026.
| Body | Policy Area | Current Position | Status | What to Watch |
|---|---|---|---|---|
| DfE | Generative AI in education | The August 2025 policy paper supports safe and effective use in schools and colleges. Separate product safety standards, updated January 2026, are aimed at edtech developers and suppliers; schools and colleges may use them when evaluating products. These are guidance, not a four-rule statutory code for every training provider. | Current guidance | Use the live DfE technology collection and the product-safety page; do not rely on an old summary. |
| Ofqual | AI and regulated assessment | The April 2026 AI-malpractice advice note explains how existing Conditions apply; it creates no new regulatory requirements. Awarding organisations must manage validity and malpractice risk. AI cannot be the sole mechanism for determining a learner's mark under current regulations. | Current advice | Follow the relevant awarding organisation's centre and malpractice instructions as well as Ofqual's framework. |
| Ofsted | AI during inspection | Ofsted evaluates the quality, safety and impact of a provider's practice through the applicable inspection toolkit; it does not award an AI-compliance badge. Its research reports both useful practice and governance risks. Inspection-material updates published in June 2026 take effect from September 2026, not before. | Date-sensitive | Use the current FE and skills toolkit for an inspection before September, then the published September 2026 version after commencement. |
| Skills England | AI skills products | Confirmed products include ST1512 Artificial Intelligence and Automation Practitioner and the Level 5 units AU0009, AU0010 and AU0011. AI Skills Boost separately lists free partner courses benchmarked against six foundation skills. Benchmarking or hub listing is not provider accreditation and does not make a course levy-funded. | Live products | Check the exact product page, provider permission and start-date funding rules; AU0002 is withdrawn. |
| UK regulation | AI used at work or in education | Existing data protection, equality, consumer, employment, intellectual-property, safeguarding and sector rules can apply to AI now. Government's principles-based policy and regulator activity should not be described as a single universal “UK AI Act”. Check the law and regulator for the actual use case. | Use-case specific | Track primary legislation and regulator guidance; do not publish a forecast consultation date as settled law. |
| ICO | AI and personal data | All DUAA data-protection provisions are now in force. The automated-decision rules are broader than the old regime but retain safeguards, with tighter restrictions for special-category data. A DPIA is required where processing is likely to result in high risk, not automatically for every tool solely because it uses AI. | Law and guidance | Use the ICO's current AI guidance and DUAA change summary; document lawful basis, transparency, necessity, processors and human challenge routes. |
| JCQ | AI in assessments | JCQ's revision-two guidance is effective from 30 April 2025 for assessments within its scope. Learners must not submit AI-generated work as their own and must acknowledge permitted use. It is not a blanket ban on every AI tool in every qualification; centres must apply the specification and awarding-body instructions. | Current guidance | Check the current JCQ document and each awarding organisation's rules before giving learner advice. |
| Jisc and ETF | Sector practice and CPD | Jisc and the Education Training Foundation publish useful sector resources and professional development, but these are not regulators. Do not present a course, research report or maturity model as a statutory “AI in Teaching Framework” unless the exact current source says so. | Non-regulatory | Verify the title, date, audience, availability and funding of any resource directly before citing it. |
Section 2: Primary Sources to Use
Open the live source and check its audience, effective date and legal status before turning it into a policy requirement.
DfE: Generative AI in Education and Product Safety
The policy paper covers opportunities, safe and effective use and responsible practice in schools and colleges. The separate product-safety standards are written for developers and suppliers, while schools and colleges may use them in procurement. Read these alongside safeguarding, data-protection and cyber guidance applicable to the setting.
DfE technology and AI collection · Generative AI product safety standards
Ofqual: AI Malpractice and Assessment
The advice note explains how existing Conditions of Recognition and guidance apply to learner misuse of AI. It explicitly says it creates no new regulatory requirements. Centres should use their awarding organisation's instructions; Ofqual also provides 2026 coursework resources and confirms that AI cannot be the sole marker under current regulations.
Ofqual AI malpractice advice note · Ofqual guide for schools and colleges 2026
Ofsted: FE and Skills Inspection Materials
The collection contains both the materials currently in force and clearly labelled September 2026 replacements. Use the version in force on the inspection date. Ofsted's AI research is useful context, but a research case study is not an additional inspection standard.
FE and skills inspection toolkit, guides and information · Ofsted AI early-adopter research
Skills England and DSIT: AI Skills Routes
Use Skills England for ST1512 and the approved unit pages. Use the AI Skills Boost explainer and foundation-skills benchmark for the free partner-course initiative. Keep these routes separate: a benchmarked course is not automatically an apprenticeship-service product or a general provider accreditation.
ST1512 AI and Automation Practitioner · AU0009 · AU0010 · AU0011
ICO: DUAA, AI and Automated Decisions
The DUAA changed the UK data-protection framework, including solely automated significant decisions. Organisations still need safeguards, and special-category data remains more restricted. Apply the ICO's high-risk test for DPIAs instead of treating every AI purchase as automatically requiring one.
ICO DUAA data-protection summary · ICO artificial-intelligence guidance
JCQ: AI Use in Assessments
The current document explains centre responsibilities, acknowledgement, marking, prevention, identification and reporting for assessments within its scope. Use it with the relevant specification and awarding-organisation instructions; do not convert it into a blanket ban covering unrelated provision.
Section 3: Provider Action Checklist
This is a risk-based operating checklist, not a claim that every item is mandated by every body above.
Classify the source and scope
- Record whether each source is legislation, statutory guidance, regulatory conditions, non-statutory guidance, an inspection toolkit, research or a funding rule
- Record its jurisdiction, audience, publication date, effective date and next review owner
- Use the inspection and funding version in force on the relevant date
Govern personal data and automated decisions
- Describe each AI use case, personal data, purpose, lawful basis, users, supplier, retention and data location
- Complete a DPIA where the processing is likely to result in high risk and document the screening decision where it is not
- Identify solely automated decisions with significant effects, special-category data and the applicable DUAA safeguards
- Give people meaningful information and a practical route to obtain human intervention, make representations or contest a qualifying decision
- Put processor, security, deletion, model-training and incident obligations into supplier terms
Protect assessment integrity
- Map the exact awarding organisation, assessment plan or JCQ instruction to each programme and assessment type
- Tell learners what use is allowed, what must be acknowledged and what constitutes malpractice
- Design authentic assessment and verification rather than relying on an AI-detection score as proof
- Train assessors and IQA staff on evidence, escalation, fairness and record keeping
- Do not use AI as the sole marker where Ofqual's regulations apply
Procure and operate safely
- Test accuracy, bias, accessibility, safeguarding, security and human oversight against the actual users and context
- Where the product is for schools or colleges, use DfE's current product-safety standards as part of due diligence
- Set allowed and prohibited use, review frequency, accountable owner, complaints route and shutdown criteria
- Monitor impact on learners and staff, not only adoption or time saved
- Keep a dated source register so policy wording can be changed without rewriting historical learner evidence
Four claims this tracker does not support
- “DfE requires every independent training provider to publish the same AI policy.”
- “Every AI tool always requires a DPIA.”
- “Ofsted has a standalone AI compliance grade.”
- “Any AI literacy course or AI Skills Boost listing is Growth and Skills Levy funded.”
Section 4: Related Guides
See how TIQPlus supports compliant AI-powered training delivery
TIQPlus is built to meet DfE, Ofsted, and Ofqual expectations — with evidence management, learner progress tracking, and governance features designed for the UK regulatory context.
Book a free demo