AI literacy is not a course certificate to file for the auditor. Article 4 is the cheapest change programme you will ever get funded, so use the legal requirement to get AI actually used.
Short answer: since 2 February 2025, providers and deployers of AI systems must take measures to support the development of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf. That is Article 4 of the EU AI Act (Regulation (EU) 2024/1689, EUR-Lex). Mid-2026, the Digital Omnibus on AI (Regulation (EU) 2026/1744) replaced Article 4: the obligation remains, but it does not require guaranteeing any specific or 'sufficient' level of literacy for any individual (European Commission, AI literacy – Questions & Answers). It does not only apply to AI companies. In practice it applies to any employer whose people use AI at work.
What Article 4 actually says
The article is short. It requires measures that take into account staff's technical knowledge, experience, education and training, the context in which the AI systems are used, and the people they are used on. There is no fixed curriculum, no minimum number of hours and no certification requirement. After the Digital Omnibus clarification, providers and deployers must support the development of AI literacy — they do not have to guarantee any specific level for any individual. National market surveillance authorities enforce Article 4 from 2 August 2026 (European Commission, AI talent, skills and literacy).
The European Commission's AI literacy Q&A confirms that no specific certificate is required, that there is no obligation to measure each employee's knowledge, and that measures should be proportionate to the organisation's role and risks (European Commission, AI literacy – Questions & Answers).
In other words, the law tells you to support development, not to certify a 'sufficient' level. That is exactly where most organisations take the easy route.
What most answers say
Legal summaries of the AI Act focus on risk classes, prohibited systems and deadlines. AI literacy often still gets a single outdated line: "ensure staff have sufficient knowledge". In practice that becomes a 30-minute e-learning module everyone clicks through.
That may satisfy the letter of the law. It misses the opportunity.
Why AI literacy is a change question, not a compliance question
Prosci's study of 1,107 professionals found user proficiency was the single biggest AI implementation challenge, at 38% of all reported difficulties. Technical issues accounted for 16% (Prosci). At Lyon AM our view is that roughly 70% of a successful AI programme is about people and ways of working.
So the work the law requires (building understanding of how AI works, its limits and how to use it properly) is the same work that decides whether your AI investment pays back. You can do it as a paper exercise, or as the first step in real change.
Building AI literacy that lasts, in five steps
1. Map who uses what
Start with an inventory: which AI systems are in use, by which roles, for which tasks? Include AI built into tools you already own, and tools employees use without approval.
2. Segment staff by role and risk
A salesperson drafting emails with AI needs different knowledge from an HR specialist using a system to screen applications. AI used in recruitment and in decisions affecting employees' terms is classed as high-risk under Annex III of the Act. That needs more.
A simple model has three levels:
All staff: what AI is and isn't, your rules, how to check outputs.
Active users: practical use in their own tasks, typical failure modes, when not to trust the answer.
Owners and managers: risk assessment, accountability, how to track usage and quality.
3. Teach on real work
Generic AI courses build general awareness, not changed behaviour. Have each group work on their own tasks. It is better learning, and it makes it easier to show your measures fit "the context the AI systems are used in", as Article 4 requires.
4. Make managers the carriers
Managers decide whether AI becomes part of daily work. Train them first and give them responsibility for follow-up in their teams.
5. Document what matters
Record which groups received which training, when, and how you follow up. But also document what changed in the work itself. That is better evidence of AI literacy than an attendance list.
The Nordic context
In Sweden and the wider Nordics, the AI Act sits alongside co-determination and work environment law. If AI changes tasks or how employees are monitored, Swedish employers may need to negotiate with unions under section 11 of the Co-determination Act (MBL) before deciding (Unionen). Planning AI literacy together with employee representatives makes it easier and more credible.
Checklist: Article 4 AI literacy
Inventory of AI systems and users
Roles segmented by usage and risk
Training tailored to role and context
Clear rules (an AI policy) the training builds on
Managers trained first
Records of measures and follow-up
A plan to update as tools and rules change
Frequently asked questions
Does the AI literacy requirement apply to every company? It applies to every provider and deployer of AI systems in a professional context. If your staff use AI tools at work, you are in practice covered.
When did Article 4 start to apply? 2 February 2025. The Digital Omnibus clarification entered into force mid-July 2026; national market surveillance authorities enforce from 2 August 2026.
Do employees need a certificate? No. According to the European Commission's Q&A, no certificate is required. Measures should fit the role, knowledge and context.
Is one e-learning module enough for everyone? It can be part of it, but a single generic course rarely reflects different roles and risks, which Article 4 explicitly requires you to consider.
Next step
Start by seeing where you stand. The Lyon AM AI diagnostic shows how much AI is used and how ready your foundations are. Want an AI literacy programme that also moves the business? Contact us.
This article is general information, not legal advice.
