dimensions scored against UNESCO and OECD benchmarks
on policy framing and on student curriculum
on data protection, the weakest dimension by a distance
initial agreement between coders on the double-coded subset
India has committed to AI education about as firmly as a country can. NEP 2020 mandates AI competencies across schooling, NITI Aayog's strategy sets the national direction, and CBSE has built a full curriculum for Grades IX to XII.
What has not existed is a structured account of how that commitment measures against what international bodies actually recommend, and where it falls short.
That is the gap this study fills, dimension by dimension.
Qualitative document analysis of the national corpus: NEP 2020, NITI Aayog's National Strategy for AI, the CBSE AI curricula, the AI for All initiative, and NCERT and DIKSHA resources.
Six dimensions, derived from UNESCO and OECD guidance and refined during reading: policy framing, ethics and governance, teacher development, student competencies, data protection, and implementation mechanisms. Each scored on a 1 to 5 rubric, where 5 is full alignment.
Around a third of documents were double-coded, giving 80% initial agreement, with the remainder settled by consensus.
Two dimensions score full marks. Three sit at partial alignment. One is close to absent. The paper's own phrase for the pattern is exact: curriculum-forward, governance-light.
The strategic commitment is real and it is national, not aspirational language in a preamble. On policy framing, India meets the international expectation squarely.
The curriculum is the stronger claim. CBSE lays out a progression from foundations in Grade IX, through the AI Project Cycle in Grade X, to specialised modules in data science, computer vision and language processing, ending in capstone projects.
That degree of operational specificity is uncommon in national curricula anywhere. Most countries have the ambition and not the sequence.
Ethics appears in the curriculum as something students learn about. It does not appear as something institutions must do. The analysis found no requirement for algorithmic transparency disclosure and no requirement for independent audit of educational AI systems, both of which UNESCO and OECD frameworks call for.
Data protection is weaker still: no education-specific provisions covering AI tools, no standards for the vendors supplying them. Students are taught about digital privacy while institutional responsibility goes unassigned.
Teacher development has the delivery channels, through DIKSHA and NISHTHA, but not yet the depth or national consistency that structured professional learning implies.
Nothing here suggests India misunderstands the problem. NEP 2020 already mandates a National Educational Technology Forum to evaluate interventions before they scale and to set standards. It remains largely unoperationalised.
So the finding is narrower and more useful than "India should do better". Policy specifies what students should learn without specifying who governs the systems, who trains the teachers to a standard, or who holds the data.
Any large education system can run itself against the same six dimensions and find out which of them it has quietly left empty.
| Dimension | Score | What was found |
|---|---|---|
| Policy framing | 5 | NEP 2020 and the AI for All strategy, operationalised through CBSE curricula |
| Student competencies | 5 | CBSE AI curriculum, Grades IX to XII, with a defined project cycle |
| Ethics & governance | 3 | Ethics taught as curriculum content; no oversight body, no algorithmic accountability |
| Teacher development | 3 | DIKSHA and NISHTHA modules exist; not systemic, not credentialed |
| Implementation | 3 | Distribution runs on existing structures; no dedicated authority |
| Data protection | 1 | No education-specific data protection, no standards for AI tools |
The study is a reading of intended strategy, so it draws on existing empirical work to show how wide the gap to practice runs. Large-scale surveys of Indian teachers report that more than 70% now use AI tools, and that around 60% use them for lesson planning, while only 57% correctly identified a basic misconception about AI.
High usage alongside shallow understanding is not a training-attendance problem. It is what happens when delivery exists but depth and credentialing do not, which is precisely what a score of 3 on teacher development describes.
This is an analysis of policy documents and curricular frameworks. It describes intended strategy, not what is happening in classrooms, and the two are known to diverge sharply in India, where adoption is uneven, often confined to substitution-level tasks, and mediated by regional and infrastructural disparity. The analytical boundary is national K to 12 programmes, meaning CBSE, NEP 2020 and NITI Aayog. State-level AI policies and higher education initiatives are deliberately out of scope, and a fuller picture would need them.
Shenoy, P., & Saarela, M. (2026). AI literacy and governance in Indian education: Mapping policy approaches and institutional initiatives. In 2026 International Conference on Artificial Intelligence, Computer, Data Sciences and Applications (ACDSA) (pp. 1–6). IEEE.
Sessions and seminars on this work usually run to about ninety minutes. The shape of one is set after a training needs analysis with the school or organisation's own leadership, never before it.