Blog · Training

Mandatory AI training for employees: what the AI Act really requires

Since February 2025, every organisation using AI must ensure its teams' AI literacy. Who is concerned, what 'literacy' means, and how to comply without a heavy file.

By Christophe Delebarre · 28 October 2026 · 4 min read

Article prepared with the help of artificial intelligence from our monitoring, then reviewed and approved by Christophe Delebarre before publication.

The word "mandatory" is frightening, so let us start with the text. Article 4 of the European regulation on artificial intelligence, applicable since 2 February 2025, requires providers and deployers of AI systems to take measures to ensure, "to their best extent", a sufficient level of AI literacy among their staff and other persons using those systems on their behalf. Deployer means you: any company, association or local authority that uses an AI system in the course of its activity.

Who is concerned

Almost everyone, and sooner than you think. A writing assistant used for quotes, a meeting transcription tool, a chatbot on the website, a sorting module in the mailbox: these are AI systems within the meaning of the regulation. As soon as an employee uses one at work, the organisation is a deployer. Size changes nothing: the eight-person SME is concerned just like the group. Local authorities too.

What the text does not say, though you sometimes read it: it imposes no certification, no minimum duration, no approved body. It is an obligation of means, not of result. You must be able to show that you acted, in proportion to your uses and to your teams' profile.

What "AI literacy" means

The regulation defines it as the skills, knowledge and understanding that allow an AI system to be deployed in an informed way, and awareness of its opportunities and risks. Translated for a ten-person department, that gives three levels:

  • Understanding what an AI system is and is not: a probabilistic tool that can be confidently wrong, not a search engine nor a colleague.
  • Knowing how to use it in one's job: what to entrust to it, what to check, which data never to put in, how to report a doubtful result.
  • Knowing the framework: GDPR, transparency towards customers and citizens, human validation on the actions that matter.

The expected level depends on the role. Someone who validates reminders prepared by an assistant does not need the same training as someone who configures that assistant. The regulation says so explicitly: context of use, persons concerned, prior knowledge.

How to comply, in practice

Four proportionate steps are enough in most organisations.

  • Inventory the uses. Which AI tools are used, by whom, for what. You often discover uses nobody had declared, and that is precisely what the regulation wants to avoid.
  • Write a one-page usage framework: authorised tools, forbidden data, verification rules, who to contact in case of doubt.
  • Train the teams, by level: a short awareness session for all, in-depth training for those who use AI daily or configure it.
  • Keep the record: list of participants, dates, contents. That is what you will show if asked.
And the penalties? Article 4 carries no fine of its own. But since August 2026, national authorities supervise the application of the regulation, and a lack of training will weigh in the assessment of another breach: a data leak through an AI tool, a contested decision, misleading content. Training your teams also reduces those risks.

Funding the training

Training delivered by a Qualiopi-certified organisation is eligible for funding by the French skills operators (OPCO). For an SME, that often means partial or full coverage of the teaching cost. The courses we designed with Manager's Solution, a Qualiopi-certified organisation, cover artificial intelligence, GDPR and the DPO role, cybersecurity and information systems management, in open sessions or on your premises, in person or remotely.

Training is not enough, and that is a good thing

Training teams in AI without giving them a framed tool is inviting them to use consumer tools with company data. The reverse is also true: deploying an assistant without training means watching it sit in a corner. The two go together, which is why we train teams before deploying an automation, never after. Our five-step method provides for it explicitly.

For course details and arrangements, see the AI training for teams page, or request an in-house quote.

Related articles