The European AI Regulation (AI Act) obliges providers and users of AI systems to ensure a certain level of AI literacy among their staff. The idea behind this is simple: anyone who works with an AI system or bases decisions on it must understand what the system does, what data it uses, and where the limits of its reliability lie. Without that understanding, no one can properly judge whether an outcome is correct or whether something is going wrong.
The exact obligations, who precisely they affect and by when they apply, are set out in the regulation itself and the accompanying implementing documents. Because texts and application dates can be amended or further specified, the current state of affairs can be found at the official EU sources and at supervisory authorities that publish on this topic. This page describes the mechanism, not the precise date or the exact article number.
AI literacy is not a diploma requirement or a course obligation in itself. It concerns a functional result: employees who work with an AI system must be able to:
The weight of this requirement depends heavily on the risk level of the system. For systems classified as high risk, as set out on the page about high-risk AI in the workplace, the expectations regarding knowledge, oversight and documentation are considerably higher than for a simple tool for text or planning.
AI literacy directly affects the question of which tasks an organisation leaves to AI, which remain under supervision and which remain human work. That is precisely the three-way division that ftetoai.com assesses per task:
1. the AI performs the task independently; 2. the AI performs the task partly, with an employee who approves or rejects it and substantiates that with a reason; 3. the task remains human work.
The second category is precisely where AI literacy is most needed. An employee who has to approve or reject an AI outcome can only do so meaningfully if he understands what the system has done and where it typically goes wrong. Without that understanding, "human oversight" becomes a signature without substance, which does not reduce the risk of errors but merely shifts it.
This dependency differs greatly per sector and per task. Wholesale trade involves different systems and risks than manufacturing or the transport sector: see, for example, the overviews of what AI can take over in wholesale trade, what AI can take over in manufacturing and what AI can take over in the transport sector. The required AI literacy follows the task, not the job title: two employees with the same job title may need a different level of knowledge, depending on which AI-supported tasks they actually carry out.
Organisations that deploy AI systems must, in practice, record and safeguard:
These points also touch on the broader data processing surrounding AI systems. When a task processes personal data, the requirements described on the page about the GDPR when automating tasks also apply, in addition to the requirements regarding AI literacy itself.
AI literacy is not a basis for personnel decisions, and this page provides no substantiation for that. Whether a lack of knowledge has consequences for a position, a role or an employment relationship is a question governed by its own legal requirements, including in the area of employment law and works council participation. This page makes no statement about that; it requires its own advice, tailored to the situation.
Nor does this page claim to know how much budget or how much training time AI literacy requires in a specific organisation. That depends on the number of systems, their risk class, the complexity of the tasks and the existing level of knowledge. Anyone looking for a figure on that will not find it in general texts but in an analysis of their own tasks and systems.
A first step is to get a clear picture of which tasks in your organisation are already carried out with AI support, and which of these are under supervision versus fully automated. That distinction immediately makes visible where AI literacy matters most.
The free quickscan from ftetoai provides an initial direction for this: twelve questions, no account required, with an indication of what proportion of the hours in your profile can be taken over by AI today. The full work scan, which goes deeper into task level and oversight requirements, is still under construction. The quickscan is an indication, not a legal assessment of the AI literacy requirement in your organisation.
Vraag maar. Ik ken de kennisbank van deze site; wat ik niet weet, zeg ik erbij.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.