Mandatory AI literacy: the EU AI Act obligation companies should already address

AI literacy is becoming a practical requirement for companies that use AI. Article 4 of the EU AI Act expects providers and deployers to take measures so staff understand AI systems appropriately for their role.

The AI Omnibus, in force since 27 July 2026, clarifies that this means supporting the development of AI literacy in context. It does not require a provider or deployer to guarantee a specific level of literacy for each individual.

For most companies, the right response is not a theoretical course. It is practical training tied to the tools people actually use.

What AI literacy should cover

Employees should understand:

• AI can produce incorrect or unsupported answers.
• Sensitive data should not be pasted into unapproved tools.
• Outputs need verification when decisions matter.
• Some use cases require human oversight.
• The source of an answer matters.
• Company-approved tools and policies should be followed.

This is basic operational hygiene.

Who needs training

Not everyone needs the same depth. A support agent, HR manager, engineer, salesperson, and administrator use AI differently. Training should match the risk and context of each role.

People using AI with personal data, customer data, employment decisions, or compliance-sensitive content need clearer rules.

How to make it lightweight

Good AI literacy can be delivered through:

• Short practical sessions.
• Examples from the company's own workflows.
• Approved and prohibited use cases.
• Data handling rules.
• A checklist for verifying AI outputs.
• Refresher guidance when tools change.

The goal is better daily behavior, not certificates for their own sake.

The practical conclusion

AI literacy is not only a compliance task. It improves productivity because employees learn when to trust, verify, escalate, or stop.

Use the role-based literacy checklist in the AI governance toolkit for SMEs, and reinforce it with a clear AI acceptable-use policy.

Polp supports safer AI use by keeping answers tied to company sources and access controls, which makes training easier to apply in practice.

Sources:

• AI Omnibus – amended Article 4
• European Commission: AI Omnibus enters into force