AI literacy at work.
Since 2 February 2025, the EU AI Act has asked companies that use AI systems to look after their staff's AI literacy. Since 27 July 2026, Article 4 asks them to take measures to support it, without having to guarantee a specific level for any individual. What the text says, what AI literacy means, and a practical place to start: the people who approve what AI agents do.
Try it: what AI literacy means for each role
Pick a role. See the AI it works with, what those people need to know, and where to practise.
Works with
An AI agent that answers billing questions and proposes refunds, credits and subscription changes.
Needs to know
- What a refund, a credit and a cancellation really do, and which ones can't be undone
- That a request in an email or a ticket isn't an order
- When the narrower option the agent offers is the better call
Works with
An AI agent that proposes DNS records, security settings, cache purges and firewall rules.
Needs to know
- Why a deleted DNS record can break email
- What a security setting protects before anyone relaxes it
- How instructions hide in logs and form fields
Works with
Coding agents that open pull requests, merge, tag releases and change access.
Needs to know
- Why a moved version tag breaks customers' installs
- Who should get write access, and for how long
- How a README or an issue can steer an agent
Works with
AI assistants that draft replies and look up customers and orders.
Needs to know
- That a draft is checked before it goes out
- Which customer details the assistant shouldn't repeat
- How a message can be written to steer the assistant
Works with
Deciding which AI systems the team uses, and who approves what they do.
Needs to know
- Which actions should always wait for a person
- Who is trained to approve them
- A record of who trained, for your files
What Article 4 says
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, taking into account their technical knowledge, experience, education and training and the context the AI systems are to be used in, and considering the persons or groups of persons on whom the AI systems are to be used. This obligation does not require providers or deployers to guarantee any specific level of AI literacy of any individual.
Article 4(1) of the AI Act, Regulation (EU) 2024/1689, as amended by Regulation (EU) 2026/1744
That's the text in force since 27 July 2026, when the Digital Omnibus on AI changed it. From 2 February 2025 until then, the article asked providers and deployers to take measures to ensure, to their best extent, a sufficient level of AI literacy.
The same article now also says that the Commission and the Member States support these efforts, in particular those of SMEs, and that the Commission publishes practical examples of how to meet the obligation.
What AI literacy means
skills, knowledge and understanding that allow providers, deployers and affected persons, taking into account their respective rights and obligations in the context of this Regulation, to make an informed deployment of AI systems, as well as to gain awareness about the opportunities and risks of AI and possible harm it can cause
Article 3(56) of the AI Act
In plain words: the people who use AI at work should understand what the systems they use can do, where they go wrong, and what that could cost.
Who it concerns
Providers build AI systems; deployers use them under their own authority, at work. A company whose staff use an AI assistant, or let an AI agent work in billing, on the website or in code, is a deployer. Article 4 is about its staff and the other people who use AI systems on its behalf.
A practical place to start
Start where a mistake by an AI would cost the most: with the people who approve what AI agents do. An agent proposes a refund, a DNS change or a merge in seconds; the person who approves it needs to know what would really happen, who asked, and what a manipulated agent looks like.
That's what Human Queue Training teaches, with real-world scenarios modelled on Stripe, Cloudflare and GitHub. It's free, in English, German and French, and an organisation can follow who trained and passed on one dashboard.
What we can and can't say
This page isn't legal advice, and no course makes a company compliant on its own. Article 4 leaves the measures to each provider and deployer. Training the people who approve your AI agents can be one of them; for what else fits your company, ask a legal adviser.
The AI Act's text from the European Commission's AI Act Service Desk (consolidated version as at 27 July 2026), checked on 11 November 2026.