UNESCO's Generative AI Guidance: A Checklist for Ministries

UNESCO's September 2023 guidance on generative AI: seven steps for governments, an age 13 minimum for chatbots, and a gap check for ministries and boards.

Guide: AI Policy and EthicsLast verified: October 20267 min read

UNESCO's Guidance for generative AI in education and research, published on September 7, 2023, is the most detailed international checklist for education ministries on the subject. It proposes steps for governments, accountabilities for tool providers, and duties for schools, and it recommends 13 as the minimum age for independent use of general-purpose chatbots. It is guidance, not law, and in Canada and the United States the binding rules sit with provinces, states, and districts.

Key takeaways

  • A UNESCO survey of slightly over 450 schools and universities in May 2023 found that fewer than 10 percent had formal guidance on generative AI.
  • The guidance asks governments to act on data protection, national AI strategies, ethics rules, copyright, AI-specific regulation, teacher capacity, and long-term effects.
  • It sets 13 as the minimum age for independent chatbot conversations and asks countries to decide whether self-reported age is enough.
  • In 2023 only seven countries reported teacher training frameworks or programs on AI, according to the guidance.
  • Because the text dates from 2023, each recommendation needs to be checked against current law and tools before it is adopted.

Why a ministry-level guide was written

When ChatGPT reached 100 million monthly users in January 2023, most education systems had no written position on it. UNESCO surveyed slightly over 450 schools and universities from its Associated Schools and university Chairs networks between May 4 and 19, 2023. Fewer than 10 percent had developed institutional policies or formal guidance. Universities were ahead of schools, at about 13 percent versus about 7 percent, and roughly 40 percent of the institutions that did have guidance said it was communicated only orally. Only two reported a ban. Close to 20 percent of respondents did not know whether their institution had any policy at all.

The gap between fast adoption and slow rule-making is the stated reason for the guidance. Its authors, Fengchun Miao and Wayne Holmes, wrote it for ministries, tool providers, and institutions at the same time, which is why the text splits duties among four groups.

The steps for governments, in plain terms

The guidance draws on a UNESCO review of national AI strategies from April 2023. It describes the result as a series of six steps, yet the text numbers seven. The table restates each one in a sentence.

StepWhat the guidance asksReason it gives
1. Data protectionEndorse international or regional data protection rules, or write national onesModel training collected online data across many countries; 137 of 194 countries had data protection laws, but enforcement was unclear
2. National AI strategiesAdopt, revise, and fund whole-of-government strategiesAbout 67 countries had national AI strategies by early 2023, and none yet addressed generative AI
3. Ethics regulationTurn ethical principles into enforceable rulesOnly around 20 countries had clear AI ethics regulations
4. CopyrightAdjust or enforce copyright law for training data and AI outputOnly China, EU countries, and the United States had adjusted copyright law at the time
5. Generative AI rulesWrite frameworks specific to generative AIAs of July 2023 only one country, China, had issued such regulation
6. Capacity buildingTrain teachers, researchers, and institutionsInstitutions need to understand benefits and risks before they can validate tools
7. Long-term effectsHold open public debate on knowledge, teaching, and assessmentEffects on education were only beginning to appear

Every figure in the right-hand column is a 2023 count reported by UNESCO and has likely changed since. The table is useful as a structure, not as a current status report.

The age line and the question of verification

The guidance says the minimum age for independent conversations with general-purpose chatbots should be 13. It traces that number to the United States Children's Online Privacy Protection Act of 1998, notes that many commentators consider 13 too young and have argued for 16, and records that the European Union's data protection regulation specifies 16 for social media use without parental permission. At the time of writing, ChatGPT's terms required users to be at least 13 and required permission from a parent or guardian for users under 18.

The more demanding part is what follows the number. Countries are asked to decide whether self-reporting of age remains acceptable, to hold providers accountable for age verification, and to define the role of parents or guardians in monitoring younger children's conversations. A ministry that adopts only the number has adopted the easy half.

Duties the guidance gives to providers and to schools

Providers are assigned ten categories of accountability, among them evidence of trustworthy data sources, labeling of generated content, clear statements of limits, complaint channels, and cooperation with regulators on unlawful use. Institutions are asked to audit the algorithms, data, and outputs of tools they adopt, to validate that a tool does no predictable harm to children and vulnerable groups, to evaluate long-term effects on skills such as critical thinking, and to consider minimum ages for independent use.

On teacher preparation, the guidance cites 2023 survey data showing that only seven countries (China, Finland, Georgia, Qatar, Spain, Thailand, and Turkiye) reported frameworks or training programs on AI for teachers. It names Singapore's education ministry as the only one that had built an online repository centered on ChatGPT in teaching and learning.

A gap check for a ministry, board, or district

The following questions turn the guidance into an audit that a policy team can run on its own documents. Each one is answerable with a file, a statute citation, or a procurement record.

  • Is there a written position on generative AI that students, teachers, and parents can read, and does it say who approves new tools?
  • Does the procurement process ask providers where training data came from, how personal data is stored, and how the tool labels its output?
  • Is a minimum age stated, and does the document say how age is checked and who is responsible?
  • Is there a named person or body that audits tools after adoption, not only before?
  • Do teachers have paid time for training, and is it required or optional?
  • Is there a complaint route for students and parents, with a response deadline?
  • Is there a date on which the policy will be reviewed against current law and current tools?

Where the guidance stops

The guidance cannot bind anyone. In Canada, section 93 of the Constitution Act, 1867 gives each provincial legislature the exclusive power to make laws in relation to education, so ministries in the provinces and territories decide how much of it to adopt. In the United States, a federal finding in 20 U.S.C. 3401 reserves primary public responsibility for education to the states and local school systems. A federal or international text can supply vocabulary and a checklist, and local law supplies the force.

It also predates many later developments, including newer tools and the education guidance several jurisdictions have issued since. UNESCO's page for the document shows an update on January 16, 2026, but the 2023 counts quoted above remain 2023 counts. This article is general information, not legal advice, and a board or ministry should confirm current statutes with its own counsel.

FAQ

Is the UNESCO guidance a law or a treaty?

No. It is a non-binding policy document. Governments can adopt parts of it into their own law, but nothing in it creates an obligation on its own.

Does it say students under 13 may never use AI?

No. It recommends 13 as the minimum age for independent conversations with general-purpose chatbots. Supervised classroom use led by a teacher is a different situation that the age line does not address directly.

Does it recommend banning generative AI in schools?

It lists banning as one response some governments had considered, alongside assessing needs and writing new rules, and it does not endorse a blanket ban. It asks for validation of tools against local ethical frameworks and for regulation.

Who in Canada can act on it?

Provincial and territorial ministries of education and, below them, school boards. The federal government has only a limited role in schooling.

Last verified: October 2026

Sources

  • Guidance for generative AI in education and research, Miao F, Holmes W, UNESCO, September 7, 2023, page updated January 16, 2026
  • UNESCO survey: less than 10% of schools and universities have formal guidance on AI, UNESCO, June 1, 2023, updated September 18, 2024
  • Constitution Act, 1867, section 93, Justice Laws Website, Government of Canada
  • Department of Education Organization Act, 20 U.S.C. 3401, findings on state and local responsibility, GPO

New to artificial intelligence?

Start with the basics, then follow the guides to tools, rules, and effects on society.

Start with AI Basics