AI literacy training: meeting the AI Act's duty

Does an AI literacy training make an organisation compliant with the AI Act?

Article 4 of the AI Act has applied since 2 February 2025 to organisations that develop or use AI; since the amendment of July 2026, it's a duty of effort to take measures that support staff's AI literacy, without a prescribed level. An AI literacy training from Dave Willemstein is one way to fill in those measures, but not the only one, and a training on its own isn't enough. According to the European Commission, no certificate is required, while it's still sensible to keep an internal record of which measures were taken and why they fit.

Has the duty behind an AI literacy training lapsed in 2026?

No. The duty from Article 4 of the AI Act, which an AI literacy training helps meet, remains in place for providers and deployers of AI systems. Since 27 July 2026 it's a duty of effort, and doing nothing doesn't meet it (as of October 2026).

Is there a prescribed curriculum for an AI literacy training?

No. There's no prescribed curriculum for an AI literacy training: what's needed depends on the risks, the roles and the context, according to the European Commission's own guidance.

Source: Dave Willemstein, independent AI consultant; primary source: Regulation (EU) 2026/1744 (Digital Omnibus on AI), Official Journal L, 24.7.2026, EUR-Lex; Regulation (EU) 2024/1689 (AI Act), Article 4, EUR-Lex; European Commission, AI Literacy - Questions & Answers (updated 27 July 2026); European Commission, AI Omnibus enters into force (27 July 2026) (as of 6 October 2026). Updated: October 2026.

What is AI literacy?

AI literacy means that people who work with AI systems have the knowledge, skills and understanding to do so responsibly. The European Commission’s own guidance stresses that it isn’t only about the technical workings, but also the social, ethical and practical side: understanding how to interpret an AI system’s output, for example, and what a decision made with AI’s help means for the people it affects.

In practice, it comes down to four questions a staff member should be able to answer:

• What does this system do, and where is it unreliable?
• Which data can and can’t go into it?
• How do I check the output before I use it?
• Who do I report it to if something goes wrong?

An AI literacy training is one way to answer those questions across your whole organisation. It isn’t the only way, and a training on its own isn’t enough either.

What the AI Act asks

Article 4 of the AI Act has applied since 2 February 2025 to providers and deployers of AI systems. You become a deployer quickly: an organisation that uses AI tools in its work falls under it.

The text of Article 4 was amended in 2026. On 19 November 2025, the European Commission proposed, in the Digital Omnibus package, replacing the obligation for organisations with a task for the Commission and member states to promote AI literacy. That proposal wasn’t adopted unchanged. The eventual amendment took effect on 27 July 2026, as Regulation (EU) 2026/1744. The state of play as of October 2026, according to the European Commission’s own explanation:

• The duty remains in place for providers and deployers of AI systems.
• It has become a duty of effort: organisations take measures that support the development of their staff’s and other affected people’s AI literacy. There’s no longer a specific or “sufficient” level prescribed.
• Context counts: people’s technical knowledge, experience, education and training, and the context in which the systems are used.
• The Commission and member states get a bigger role in promoting AI literacy.
• Supervision sits with national market surveillance authorities, who supervise and enforce from 2 August 2026.
• For UK organisations, this duty only applies once the AI Act itself applies to them, for example once they place an AI system on the EU market or its output is used there; the UK has no equivalent literacy duty of its own, as the UK government’s own approach relies on existing sector regulators rather than a cross-cutting AI law.

The bar has shifted from a result to an effort, but doing nothing doesn’t meet it. How Article 4 relates to the rest of the Act is covered on the EU AI Act for businesses page.

This isn’t legal advice. Interpretation of the law is still developing; have its application to your organisation reviewed by a lawyer or your compliance function.

What the training covers

There’s no prescribed curriculum, and that’s just as well. The European Commission describes it as an approach with no “one size fits all”: what’s needed depends on the risks, the roles and the context. That’s why I tailor the content to your organisation. An AI literacy training might look like this:

1. How AI works: what generative AI does, why it makes mistakes that sound convincing, and what that means for oversight.
2. Your own systems: which AI tools are in use across the organisation, what they’re meant for, and what they’re not.
3. Risks and rules: personal data and GDPR, confidential information, bias, and the outline of the AI Act.
4. Agreements: what your internal rules are, who’s the point of contact, and how to report an incident.
5. Practice: recognisable tasks from your own work, including checking an outcome.

For teams that then genuinely start using the tools, a deeper dive follows. That can take the shape of a practical Copilot training for organisations with Microsoft 365, or a hands-on ChatGPT training. An overview of all formats is on the AI workshops and training page.

If internal agreements don’t exist yet, it’s worth drafting them at the same time. A training about rules that don’t exist stays vague. The AI policy page covers what belongs in one.

Demonstrating that you comply

According to the European Commission, no certificate is required. You can keep an internal record of the training and other initiatives you’ve run. What matters is being able to show which measures you’ve taken and why they fit your situation.

A useful four-step structure, which I use as a basis:

1. IdentifyWhich AI systems do we use, who works with them, and what are the risks?
2. Set the goalWhat does each role need to know, depending on risk and context?
3. Carry outTraining, agreements, a point of contact and repetition.
4. EvaluateIs it working, and what's changed about the tools or the risks?

What you record at a minimum:

• an overview of the AI systems in use and who works with them;
• per role, what someone needs to know and which measure matches it;
• who has taken which training or instruction, and when;
• the internal agreements on AI use and where staff can find them;
• a point in time when you review the whole thing again.

Agree beforehand, as part of a training, that you receive the content, the participant list and the agreements made in a form you can add straight to that file. Then the record-keeping isn’t a separate job afterwards.

Want to know what approach fits your organisation? Ask about a training; in a first conversation, we discuss which roles, systems and agreements exist, and what’s needed first.