Key points
- Article 4 applies to all companies that provide or deploy AI systems, regardless of size and risk category.
- It requires measures to promote AI literacy, not a specific level for every individual.
- Good measures are role-specific: production, sales, engineering, and management need different things.
- Document what you do. That helps with the authorities and internally.
The EU AI Act is often associated with high-risk systems and prohibited practices. For most industrial companies, though, a short provision matters more day to day: Article 4 on AI literacy. It affects virtually every company that uses AI.
What Article 4 requires
Providers and deployers of AI systems must take measures to promote the AI literacy of their staff. This also applies to other persons who deal with AI systems on their behalf, such as service providers. A “deployer” is any company that uses an AI system under its own authority, even if it only purchases the system.
The obligation has applied since 2 February 2025, regardless of company size and risk category. The Digital Omnibus (Regulation (EU) 2026/1744) clarified the wording: it is about measures to promote AI literacy, not a guaranteed level of competence for each individual. The obligation itself remains. Oversight lies with the national market surveillance authorities.
What this means for industrial companies
AI is often already at work where you might not expect it: in visual quality inspection, in predictive maintenance systems, in CAD and ERP software, in assistants for writing and translation. The first step is therefore an inventory: Which AI systems are in use, who uses them, and for what?
Role-specific, not one-size-fits-all
Mandatory training for everyone with the same slides rarely serves the purpose. A better approach is a basic module for everyone plus deeper content for each role:
- Executive leadership and management: opportunities, risks, responsibility, and guardrails.
- Engineering and IT: how it works, limitations, data quality, integration.
- Procurement, sales, and service: sensible use cases, checking outputs, protecting confidentiality.
- Production: What does the system decide, what doesn’t it decide, and when does a human step in?
- External parties: service providers that operate AI systems on your behalf.
A pragmatic roadmap
- Build an AI inventory: systems, user groups, purposes.
- Form groups: Who uses what, how intensively, and with what risk?
- Define measures: a basic module for everyone, deep dives for each role, formats ranging from workshops to short learning videos.
- Publish a policy: Which tools are approved, which data is off-limits, and who do people ask?
- Document: Who completed which measure, and when?
- Refresh: when new systems or roles come in, and at regular intervals.
More than an obligation
AI literacy is the precondition for AI to pay off at all. Trained employees use the tools more productively, spot errors faster, and are less likely to turn to unapproved tools. The obligation under Article 4 is therefore a good reason to tackle the topic in a structured way.
Bottom line
Article 4 doesn’t require a certificate for every person, but it does require a serious, verifiable approach to the topic. Companies that train by role and document what they do meet the obligation and, at the same time, prepare their business for the productive use of AI.
This article is for general information only and does not constitute legal advice.
Facing a similar question?
We listen, put things in context and tell you honestly whether and how we can help.