Article 4 of the EU AI Act requires providers and deployers of AI systems to take measures to support the development of AI literacy among their staff and anyone using AI on their behalf. It has been applicable since 2 February 2025. Since August 2026, national market surveillance authorities have been supervising it. In short, the duty is live and so is the checking.
Most businesses are deployers, not developers, and that is the point people miss. If your team uses AI tools at work, Article 4 reaches you, regardless of whether you ever build a model. It applies to the smallest consultancy and the largest enterprise alike.
Providers and deployers of AI systems shall 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.
The wording matters. The Digital Omnibus (Regulation (EU) 2026/1744, in force since 27 July 2026) reworded Article 4 from "ensure, to their best extent, a sufficient level of AI literacy" to "take measures to support the development of AI literacy", and the amended duty expressly does not require you to guarantee any individual's level of literacy. What it asks for is real, documented measures, judged in context. It is a standard about what you do, not a score your people must hit.
There is no headcount carve-out. A six-person agency using AI to draft campaigns carries the same literacy duty as a multinational. What changes with size is what a proportionate programme looks like, not whether the duty applies.
Who Exactly Must Train Staff?
Scope of the duty
Deployers, not just developers
The Act binds two roles: providers (who build AI systems) and deployers (who use them under their authority). Article 4 applies to both. Crucially, the duty reaches beyond your payroll. It explicitly covers other persons dealing with the operation and use of AI systems on your behalf, so a freelancer or agency using AI to deliver your work falls inside your obligation to evidence their literacy.
Does Article 4 apply to you?
Quick test: if a person relies on an AI output to do part of a job for your business, your Article 4 measures need to reach that person.
What Counts as Compliant AI Literacy?
What good measures look like
Role-relevant, risk-aware, documented
The Act does not prescribe a curriculum, a number of hours, or an approved provider. Your measures are judged against context: the technical knowledge of your people, how the AI is used, and who it affects. The EU AI Office has clarified that a single onboarding video does not satisfy Article 4, and that you should keep documentation of the measures you put in place.
In practice, a defensible programme is role-based. The right measures for an executive drafting board papers with AI are not the same as for a support agent handling calls, which are not the same as for a developer building AI features. Each role has its own risks and its own judgement calls.
What a compliant programme covers
How the AI works
Capabilities and limits of the tools staff actually use
The risks
Bias, errors, data protection, when not to trust an output
Their obligations
What the Act expects of each specific role
Safe use in practice
Human oversight and escalation for real workflows
Does It Have Teeth?
Enforcement and exposure
Supervision live since August 2026
Article 4 carries no dedicated fine of its own. Its force is indirect and real. Since August 2026, national market surveillance authorities have supervised the duty, and a gap in basic literacy weakens your position on every other requirement of the Act. Regulators assessing a later incident will look first at whether literacy measures were ever in place. An untrained workforce is a compounding liability, not a single tick-box.
Free tool: not sure whether your training would hold up? Run the AI Readiness Check to see your exposure in about ten minutes.
Compliance Checklist
Article 4 readiness
Click to check offCommon Questions
Any provider or deployer of AI systems must take measures to support the development of AI literacy for staff and others using AI on its behalf. In practice that is any EU organisation whose people use AI tools at work.
Article 4 has been applicable since 2 February 2025. Supervision by national market surveillance authorities began in August 2026.
Role-relevant training that covers how the AI works, its risks, the staff member's obligations, and safe use, backed by documentation. A generic 'what is AI' video on its own does not count.
Yes. The duty covers other persons dealing with AI systems on your behalf, so contractors and agencies using AI to deliver your work fall within your obligation.
What To Do Right Now
Ready to get your team compliant?
Staff AI Compliance Awareness · €39/person · Certificate included · 14-day guarantee