How to prove AI literacy under the EU AI Act

The short answer

You can demonstrate your AI literacy measures with internal records of training and guidance. Article 4 does not mandate a certificate, accreditation or specific training format; the Commission's Article 4 Q&A says internal records can be sufficient documentation.

What you need to keep

A proportionate evidence set commonly includes four types of documents:

1. Training record log

A spreadsheet showing who was trained, when, what was covered, and the format. Include name, role, module, date, format (live, recorded, self-paced), trainer, and refresher due date.

2. AI-use policy

A written document covering approved tools, prohibited uses, data rules, output review expectations, and incident reporting. This shows you've thought about how AI is used in your organization.

3. Training materials

Keep copies of the slides, handouts, or videos used in training. You don't need a polished course — a slide deck presented in a team meeting counts.

4. Evidence of role-based approach

Show that different roles received different content. For example, HR got a module on AI in hiring, marketing got a module on labeling, developers got a module on AI limitations and hallucinations.

Questions a supervisor may ask

The exact supervisory approach depends on the competent national authority and context. Useful questions to prepare for include:

1. Can you show who was trained and when?
2. Is the training role-based?
3. Do you have a policy?
4. Do you plan to keep it current?

They will not ask for a certificate because none is required.

What you don't need

• No certificate or accreditation
• No external training provider
• No learning management system
• No specific number of training hours
• No submission to any authority

A fast way to build the evidence set

The Literacy Pack (€99) includes all four document types — training record log, AI-use policy, training decks, and a rollout checklist — plus a staff FAQ. You can establish a practical baseline in one afternoon, then adapt it to your roles, systems and risk context.