There is a shift in how we are expected to work with AI. It’s a simple expectation, now written into law: that the people using AI actually understand what they are doing. It is a duty of AI literacy, it lives in Article 4 of the EU AI Act, and from August 2026 it is being enforced.

If you run an organisation, this is worth understanding.

What Article 4 asks of you

Article 4 introduces a duty of AI literacy. If your organisation uses AI, you are expected to make sure your staff understand it well enough to be able to see where it helps, and where it can cause harm.

The wording is “a sufficient level of AI literacy” among staff and anyone operating AI on your behalf. That is the whole intention, not to cause concern, but to make sure the people using these tools are capable and aware, rather than quietly guessing.

Who the AI literacy duty applies to

It applies to providers and deployers of AI systems. Most organisation are ‘deployers’: you did not build the AI, but your team uses it. How might this look?

• Copilot in the inbox.
• ChatGPT for a first draft.
• Claude for a business plan.

It will likely be a tool inside software you already pay for, especially when it comes to Microsoft.

If that sounds like your organisation, Article 4 affects you. You can read the official text of Article 4 on the European Commission’s AI Act service desk if you want the source.

There is a UK question too. The Act can reach beyond the EU where there is a European connection, such as staff or customers there, so it is worth checking your own exposure. This is a general article, and not legal advice.

Why this matters now

The duty is not that new since it has applied since February 2025. What changed is the enforcement aspect. From 2 August 2026, national authorities can supervise and act on it, so it has shifted from “something coming in the future” to “something live now”.

No exam, no certificate: what “sufficient” means

Here is the part that reassures most people. There is no set exam, certificate, or fixed curriculum you must buy.

What matters is that the learning fits the person and their role:

• leaders need to understand the risks, the ethics, and where AI belongs
• everyday users need to know how to work with their tools safely
• everyone needs to know what should never be typed into a tool

However, generic awareness slides for a whole organisation, are unlikely to be enough or effective. Real literacy is specific to the actual role of each person.

What good AI literacy looks like

You do not have to be perfect, but you have to be able to show you have taken it seriously.

In practice that means role-appropriate training, delivered and written down – who learned what, and when. A single well-run session that you can document, counts for far more than good intentions alone.

A human in the loop

Article 4 is really asking for something I already believe in: a human firmly in the loop. People who understand their tools, use them with care, and can say honestly where they stand on ethics and privacy. It’s simply a good way to work.

Where to start with AI literacy

If AI is already in your business, and it almost certainly is, just take the first step. Look at who is using what. Ask whether they genuinely understand it. Then close the gaps in training and awareness, role by role, and keep a note of what you did.

That is the gist of Article 4.

If you would like training support with it – role-based AI training your team will genuinely benefit from, with governance and ethics built in – this is exactly the work I do.

Get in touch and we’ll talk it through, no pressure either way.

Or if you want a heavy read – https://digital-strategy.ec.europa.eu/en/faqs/ai-literacy-questions-answers