Several bodies have now published something they call an AI literacy framework. They were written for different audiences and they look different on the page. Underneath, they agree on most of what matters, and the disagreements are mainly about emphasis rather than substance.
Nearly all of them cover four things: understanding what AI is, using it, evaluating what it produces, and using it responsibly. Where they part company is weighting, and weighting follows audience.
At a glance
| Framework | Publisher | Audience | Structure |
|---|---|---|---|
| AI Act, Articles 3(56) and 4 | European Union | Providers, deployers and affected persons | A legal definition and a duty to take measures. No curriculum |
| AI Literacy Framework | US Department of Labor | Workers and workforce programmes | Five content areas, seven delivery principles |
| Empowering Learners for the Age of AI | European Commission and OECD | Primary and secondary education | Competences for engaging with, creating with and managing AI |
| AI Fluency | Anthropic, with academic partners | Individuals using AI assistants | Delegation, description, discernment, diligence |
| AI Literacy syllabus | AI Assurance Council | Working adults | Five assessed domains, with weightings |
The workplace framework
The US Department of Labor material is the most directly useful to an employer, because it was written about work rather than about learning. Its five content areas are to understand AI principles, explore AI uses, direct AI effectively, evaluate AI outputs, and use AI responsibly.
Its real contribution is the fourth. Treating evaluation of output as a content area in its own right, rather than a footnote to using the tool, is the distinction most training material misses. It also carries seven delivery principles, of which the practical ones are experiential learning, embedding the learning in real work, and designing for change rather than for a fixed syllabus.
The legal definition
The EU AI Act does not teach anything. It defines AI literacy at Article 3(56) as the skills, knowledge and understanding needed to make an informed deployment of AI systems and to be aware of the opportunities, risks and possible harm. Article 4 then places a duty on providers and deployers to take measures supporting it.
What the Act conspicuously does not do is prescribe a curriculum, a format, or a certificate. That is deliberate, and it is why any claim that a particular course or credential satisfies Article 4 should be read with suspicion. The obligation is on the organisation and it is assessed on what the organisation did.
The education framework
The Commission and OECD work is aimed at schools and is the most developed on critical thinking and creative use. It is also the least transferable to an employer, because its learner is a child in a structured programme rather than an adult with a deadline. It is worth reading for the competence descriptions and not worth adopting wholesale for staff training.
The individual-use framework
Anthropic’s AI Fluency material is built on four competencies: delegation, deciding what to hand over; description, communicating the task well; discernment, evaluating what comes back; and diligence, using AI responsibly and owning the result.
It is narrow by design, focused on one person’s working relationship with an assistant, and it is the quickest of the five to apply to your own habits this week. Note that its last two competencies are literacy concerns rather than fluency ones, which is the point made elsewhere about the two words describing overlapping things.
Where they agree
| Common theme | US DOL | AI Fluency | Council domain |
|---|---|---|---|
| Understanding AI | Understand AI principles | — | AI fundamentals, How AI works |
| Using AI | Explore uses, direct AI effectively | Delegation, description | Generative AI and everyday use |
| Evaluating output | Evaluate AI outputs | Discernment | Generative AI and everyday use |
| Responsible use | Use AI responsibly | Diligence | Risks and limits, Governance awareness |
The convergence is strong enough to be useful. An organisation that covers those four themes is substantially aligned with all of them, whichever it names in its policy.
Where they differ
Workplace frameworks weight data protection and accountability heavily, because that is where an employer’s exposure sits. Education frameworks weight creativity and societal impact, because that is what a curriculum is for. Individual-use frameworks concentrate on the exchange with the assistant and say comparatively little about governance.
The Council’s syllabus is written for working adults, which is why risks and governance together carry 40% of the marks. That is a choice about audience, not a claim that the others got their weightings wrong.
Which to use
Designing staff training, start from the Department of Labor content areas. Documenting measures under the AI Act, use the Act’s own definition as the reference point and record what you did against it. Improving your own habits, the four competencies of the fluency framework are quick to apply.
Measuring whether any of it worked is a different question from all of these, because none of the frameworks includes an assessment. The Council’s AI Literacy Certificate is free, identity-verified, and returns a score by domain, so the result says where the gap is rather than only whether there is one.