A UN body and an EU law both try to define AI literacy

Start here

‘AI literacy’ gets used loosely enough that it can mean anything from knowing a chatbot's name to understanding how a model is trained. Two documents from different kinds of institutions — a UN cultural and education body, and a piece of European law — each define the term with enough specificity to check yourself against.

What the documents say

UNESCO's AI competency framework for students, published 8 August 2024, defines competency across four domains — a human-centred mindset, the ethics of AI, AI techniques and applications, and AI system design — each assessed at three progression levels: understand, apply, create. It was built for school curricula, but states its aim as preparing ‘responsible citizens in the era of AI,’ and UNESCO separately publishes a companion framework for teachers, suggesting the underlying domains are meant to generalise past the classroom. The European Commission's DigComp framework, maintained by the Joint Research Centre since 2013 and now in its fifth edition, describes ‘what is needed to be digitally competent in today's society’ for the general adult population, across skill levels from basic to advanced. Separately, Article 4 of the EU AI Act, in force since 2 February 2025, gives the concept legal weight rather than leaving it as guidance: it requires providers and deployers of AI systems to ‘take measures to support the development of AI literacy of their staff,’ tailored to a person's technical knowledge and the context of use, though it does not guarantee any individual reaches a specific literacy level.

Check this

Take one of UNESCO's four domains and ask whether you could explain it in a sentence: what a human-centred mindset means when you use a chatbot, what ethics questions it raises, how it technically works, and how its designers made choices you did not see. A blank domain names a specific gap rather than a vague sense of not knowing enough.

What holds and what fails

What holds across both documents is that they treat AI literacy as a defined set of competencies with levels, not a single yes-or-no state a person either has or lacks. What is not yet established, because neither document measures it, is how many working adults meet any given level — UNESCO's framework was built for students, and Article 4 requires effort but not a guaranteed outcome, so the ‘sufficient level’ it names is not independently verified in these pages. Editorially, using UNESCO's four domains as a personal checklist is a reasonable adaptation, but it is an adaptation, not something UNESCO wrote for adult self-assessment.

• Rate yourself informally against UNESCO's four domains, one at a time.
• If your employer uses AI tools and operates in the EU, ask what measures it has taken to meet the Article 4 literacy obligation.
• Treat ‘AI literate’ as a set of separate skills, not one grade.

Neither document was written as a public self-test, but both give a more specific vocabulary than ‘keeping up with AI,’ which is the more useful thing a framework can offer a reader who is not a student and not a compliance officer.