Five principles for the effective use of AI in vocational education and training development

Policy brief

23 June 2026

Key messages

The use of artificial intelligence (AI) is emerging to support the development and updating of vocational education and training (VET) curriculum and qualifications (hereafter ‘VET development’). It is increasingly used for tasks such as labour market analysis, drafting and synthesising inputs across stakeholders, with potential to improve efficiency, responsiveness and alignment with evolving labour market demand. Across countries, its use remains ad hoc, experimental and focused on preparatory stages, such as analysing large volumes of labour market data, mapping skills, synthesising stakeholder feedback and producing initial drafts. While these applications can improve speed, consistency and evidence use, they also raise important questions about governance, quality, accountability and ownership in VET systems.

VET development has unique institutional characteristics to be considered when applying AI. Unlike other education contexts, VET curricula and qualifications are developed through negotiated, regulated and multi‑stakeholder processes, often embedded in legal frameworks and quality‑assurance arrangements and involving public authorities, social partners, sector bodies and providers. Decisions taken in these processes carry strong signalling effects for learners and employers and have direct implications for labour‑market outcomes and public trust in VET. Evidence shows that AI use in VET development remains uneven and shaped by different concerns, barriers and potential risks.

There is a strong need for VET-specific guidance on the effective and secure use of AI in VET development processes, as existing international AI principles – while providing an important foundation – do not fully account for the institutional realities of VET systems. Without VET‑specific guidance, AI risks being applied in fragmented ways that may weaken collective decision making, blur responsibility or reinforce divides between stakeholders with different levels of digital readiness. Clear guidance is therefore needed to ensure that AI strengthens, rather than undermines, the collaborative governance, legitimacy and quality that are central to VET systems.

This policy brief addresses this gap by setting out five principles to support VET stakeholders on the effective and secure use of AI in VET development. These principles clarify how AI should support collective expert judgement and decision making rather than replace it:

1. Human-centred use emphasises that AI should function as a support tool for VET development, assisting with analytical or technical tasks while final judgement, validation and responsibility remain firmly with human experts and stakeholders.

2. Diversity and inclusiveness focus on ensuring that AI use does not advantage only well- resourced stakeholders, but instead supports broad and inclusive participation in VET development, including SMEs, smaller providers and stakeholders with lower digital or AI capacity.

3. Accountability clarifies that responsibility for decisions, processes and outcomes in VET development must remain clearly assigned to VET stakeholders, even when AI is used, preventing ambiguity about who owns and approves curricula and qualifications.

4. Transparency requires that VET stakeholders understand when and how AI is used in VET development, for what purpose, on which data and with what limitations, helping to maintain trust supports continuous improvement, and avoid over-reliance on AI-generated outputs.

5. Data quality, security and protection highlights the need for high quality, relevant and well- governed data, alongside strong safeguards for sensitive educational and labour market information, ensuring that AI use complies with legal requirements and preserves trust in VET systems.

What’s the issue?

Copy link to What’s the issue?

AI is increasingly influencing how VET curricula and qualifications are developed, reviewed and aligned with labour-market needs. As skill demands evolve rapidly due to technological change, the green transition and demographic shifts, VET systems face growing pressure to become more agile, responsive and evidence informed. In this context, AI offers a range of innovation opportunities for VET development, including analysing large volumes of labour-market data, accelerating drafting and revision processes, supporting modularisation and micro-credentials, and supporting co‑ordination across many VET stakeholders involved in VET development. At the same time, the integration of AI into these processes raises important governance, quality and trust-related challenges.

Despite the growing body of international guidance on AI in education, these are not tailored to AI use in VET curriculum and qualification development. Existing AI principles and education and training-focused frameworks provide an important foundation (see Box 1), but do not fully reflect the institutional, legal and procedural characteristics of VET systems in details. VET curricula and qualifications are the outcome of negotiated and often regulated processes involving public authorities, social partners, sector bodies and providers, and they carry legal, signalling and quality-assurance functions that directly affect learners, employers and public trust. Evidence from the OECD work (OECD, 2026[1]) underpinning this policy brief shows that AI use in VET development remains uneven and largely experimental, shaped by concerns around accountability, ownership, data protection, quality assurance and stakeholder capacity. These features make a VET-specific approach to AI governance both necessary and timely.

Box 1. Examples of relevant international AI principles, guidelines and regulatory frameworks

Copy link to Box 1. Examples of relevant international AI principles, guidelines and regulatory frameworks

OECD AI Principles and policy guides

The OECD AI Principles (OECD, 2019[2]) underpin the five principles in this policy brief through their emphasis on human-centred values, transparency, robustness and accountability. In particular, the OECD concept of shared stewardship of trustworthy AI aligns with the need for collective ownership, clearly defined roles and institutional responsibility in VET development. Related OECD guidance on transparency, explainability, data governance and human capacity (OECD and Education International, 2023[3]; Varsik and Vosberg, 2024[4]) further informs the brief, highlighting conditions necessary for responsible AI use in education systems. The OECD Digital Education Outlook 2026 (OECD, 2026[5]) reinforces the importance of purposeful, human‑centred AI use, supported by explainability, supervised testing and AI literacy – considerations that are particularly relevant for VET development processes.

EU Artificial Intelligence Act, General Data Protection Regulation and guidelines

The European Union’s Artificial Intelligence Act (European Parliament and Council of the European Union, 2024[6]) provides a binding regulatory reference where AI systems used in education and training, including VET development, are classified as high‑risk. Its requirements on human oversight, responsibility allocation, risk management and documentation reinforce accountability and transparency. The Act’s provisions on data governance support data quality, security and protection, while its emphasis on non-discrimination and accessibility aligns with inclusiveness considerations.

In parallel, the General Data Protection Regulation (GDPR) (European Parliament and Council of the European Union, 2016[7]) establishes the legal baseline for lawful and secure data handling of personal and sensitive data used in AI‑supported VET development processes. The European Data Protection Board (EDPB) further clarifies how GDPR applies to AI through opinions and recommendations, most recently Opinion 28/2024 (December 2024), which emphasises lawful, fair, transparent and purpose‑specific data processing and AI use. EDPB guidance highlights obligations such as informing individuals about the use and impact of AI systems, maintaining records and risk assessments, applying privacy‑preserving techniques, and implementing robust security and governance arrangements to ensure ongoing compliance.

Complementing these instruments, the European Commission’s Guidelines on prohibited AI practices (European Commission, 2025[8]) clarify the scope and application of Article 5 of the AI Act. These non‑binding guidelines identify AI practices deemed unacceptable in light of fundamental rights and European values, including certain practices relevant to educational contexts such as harmful manipulation, exploitation of vulnerabilities, social scoring and certain biometric or emotion‑recognition practices. While not sector‑specific, the guidelines provide an important reference for VET authorities and providers when assessing which AI uses are impermissible and when determining appropriate safeguards during piloting or implementation.

G7 Toolkit for AI in the Public Sector

The G7 Toolkit for AI in the Public Sector (G7, 2024[9]) emphasises co‑ordinated governance, clarity of institutional roles, leadership and ethical safeguards in public-sector AI use. Its focus on practical governance arrangements and public accountability reinforces the process-oriented nature of the five principles, particularly those related to accountability, transparency and inclusiveness.

National and sector-specific regulations and guidance on AI in education

Countries are increasingly translating international AI principles into national or sector‑specific guidance for education and training. Examples include, Australia’s AI Transparency Statement (Australian Skills Quality Authority (ASQA), 2025[10]) and Australian Framework for Generative AI in Schools; the Slovak Republic’s Plan for the Responsible Use of AI in Education (Ministry of Education, Research, Development and Youth of the Slovak Republic, 2025[11]); and Ireland’s Guidance on AI in Schools. The UK-based Ethical Framework for AI in Education (The Institute for Ethical AI in Education, 2024[12]) complements the OECD principles by providing an education-specific lens. Its emphasis on ethical design, informed participation, professional autonomy, transparency and accountability aligns with the principles on human-centred use and transparency. The framework also highlights privacy, inclusiveness and purposeful AI use, stressing that AI should support clearly defined educational goals and avoid reinforcing existing inequalities.

Why is this important?

Copy link to Why is this important?

As AI becomes more integrated into VET development, clear principles are needed to guide its effective and secure use. The OECD, in collaboration with the European Commission, has developed this policy brief based on international AI principles and informed by country examples, surveys and stakeholder interviews, focusing specifically on the use of AI in the development of VET curricula and qualifications. The five principles each followed by their application in practice and relevant country examples are designed to support VET policymakers, qualification authorities, social partners and VET providers involved in these processes by clarifying how AI can be used to strengthen shared responsibility, collaboration, accountability and quality, rather than fragmenting governance or concentrating influence.

These principles address core dimensions of effective and secure AI use in VET development, including human-centred use, diversity and inclusiveness, accountability, transparency, and data quality, security and protection. Together, they form a coherent and complementary framework for managing how AI is used in curriculum and qualification development processes, rather than prescribing what AI tools should be used or what outcomes AI adoption should deliver. They do not provide technical guidance on specific AI systems or applications, nor do they replace existing legal or regulatory requirements.

Underlying all these is the need for stakeholder capability. AI can only enhance VET development when the people using it – VET teachers, curriculum developers, sector experts, employers and policymakers – have the skills, confidence and critical understanding needed to use it responsibly. Differences in digital readiness and AI literacy already shape who can participate meaningfully in AI‑supported processes. Without deliberate investment in capability-building, AI risks reinforcing existing inequalities, increasing over-reliance on automated outputs or weakening professional judgement. The five principles therefore assume that capability development is an ongoing requirement for effective AI use, closely linked to inclusiveness, ownership and accountability.

Given the rapid evolution of AI technologies and the diversity of VET systems, these five principles are designed to be applied flexibly across institutional, sectoral and regulatory contexts. Rather than serving as a static rulebook, they are intended to function as a shared reference point for dialogue, decision making and continuous learning across VET systems. Overall, they aim to help VET stakeholders leverage AI as a supportive tool for improvement, while maintaining ownership, safeguarding trust and inclusiveness, and preserving the integrity of VET curricula and qualifications.

Five principles and applications in practice

Copy link to Five principles and applications in practice

The five principles provide guidance on the effective and secure use of AI in VET development, clarifying how AI should support collective expert judgement, transparency, accountability and trust in VET systems. They were developed through an iterative process drawing on evidence from OECD surveys covering 25 countries and ten country case studies, including almost 80 stakeholder interviews, which identified common AI use cases, enabling conditions and recurring challenges across VET systems (OECD, 2025[13]; OECD, 2026[1]). These insights were synthesised with existing OECD, EU and international AI principles and governance frameworks to ensure coherence with established standards, while tailoring the principles to the distinctive, multi‑stakeholder and often regulated nature of VET development processes. The initial principles were subsequently discussed and refined through online workshops with VET stakeholders, collecting feedback from 117 participants (89 unique) across 21 countries. Relevant country experiences are drawn from case studies (OECD, 2026[14]).

1. Human-centred: Ensure that any use of AI in VET development is human-centred

Principle: AI use in VET development should be governed by collective human judgement and shared responsibility, with AI supporting decision making rather than replacing it.

In practice, this means that:

AI is used as a support tool, assisting high‑volume or technical tasks (e.g. skills mapping, overlap checks or preparing summaries and drafts), while final decisions and responsibility remain with qualified VET experts and stakeholders.
AI-supported insights are treated as inputs to discussion and negotiation, and are reviewed and validated through established governance processes (such as working groups, sector councils and stakeholder consultations), with human involvement maintained at all stages – from defining objectives and selecting data to validating outputs and approving final curricula and qualifications – to ensure accuracy, contextual relevance, legitimacy and overall quality.
The use of AI is guided by clearly identified needs in the VET development process, rather than by technological availability alone, and is focused on tasks where it demonstrably adds value. Where AI does not serve a clear purpose or offers limited benefit, not using AI is recognised as a legitimate and responsible choice.
AI is applied in ways that strengthen stakeholder ownership, capability and engagement in curriculum and qualification development – empowering VET teachers, social partners and employers to participate meaningfully – rather than replacing their roles, judgement or influence.