Leap Trends · Technology · Digital skills
Briefing
AI literacy and the obligation under the AI Act
- AI literacy
- AI Act
- Enforcement
Since 2 February 2025, Article 4 of the AI Act has required organisations that use AI to take measures for their people's AI literacy; since 2 August 2026 national supervisors can enforce it. The law prescribes no level or certificate, but it does require demonstrable steps.
This briefing has not been updated for more than three months. Check the figures against the source before you use them.
Compiled with AI from public sources, with the source for every figure. Key figures labelled Checked have been compared with the source by a person. How that works
Key figures
- 2 Aug 2026is when the supervision and enforcement rules for Article 4 took effect; the obligation itself has applied since 2 February 2025.European CommissionNot yet checked
- 72%of employees say expectations about the skills they need have shifted (BCG, nearly 12,000 respondents).Boston Consulting GroupNot yet checked
- 36%feel they have had enough upskilling.Boston Consulting GroupNot yet checked
What the sources say
Article 4 of the AI Act has applied since 2 February 2025 and is short: anyone who provides or uses AI systems must take measures to ensure that their own staff, and others working with AI on the organisation’s behalf, understand enough about it. The measures must suit those people’s knowledge and training, the context in which the system is used and the people the system affects. The law does not prescribe a fixed level. The Digital Omnibus (Regulation 2026/1744 of 8 July 2026, published on 24 July) relaxed the wording: organisations must take measures that support the development of AI literacy, and the obligation requires no particular level of anyone. The duty to do something remains.
The European Commission has explained in a Q&A what this means in practice. Measuring knowledge is not required, and nor is a certificate; an internal record of training and other guidance is enough as evidence. The obligation reaches beyond the organisation’s own staff and also covers contractors and service providers working with AI on its behalf. Supervision lies with the national market surveillance authorities, and the rules on supervision and enforcement have applied since 2 August 2026. Penalties come from national law and must be proportionate; the Commission explicitly states that an incident resulting from inadequate training or guidance of staff will weigh in that assessment.
In the Netherlands the Dutch Data Protection Authority (Autoriteit Persoonsgegevens, AP) is preparing for that supervision and published a second guidance document in October 2025. From its own call for input in April 2025, the AP concludes that AI use is growing faster than organisations’ approach to AI literacy, and that most organisations are still at the start-up stage. At a minimum, the AP expects an inventory of the AI systems in use, including their degree of autonomy; a picture of the risks; for each role, a picture of who works with which system and what that person needs to know; and the organisation’s ambitions with AI. The AP’s multi-year action plan has four steps: identify, set goals, implement and evaluate. And the AP falls in line with the Commission: an organisation that records what it does and why can show it later.
BCG shows how wide the gap is between the law and the workplace. Of nearly 12,000 employees, managers and leaders surveyed, 72 per cent say expectations about the skills they need have shifted, and 36 per cent consider the training their organisation offers adequate. Four in ten feel more mental pressure because of AI, and among leaders almost half.
TALIS 2024 shows how far education has got. This is the OECD study carried out in the Netherlands by the Kohnstamm Instituut and partners, which was sent to the Dutch House of Representatives on 7 October 2025. In the past year 26 per cent of primary teachers and 37 per cent of secondary teachers used AI, mainly to prepare lesson plans. Among primary teachers, 88 per cent say they lack the knowledge and skills to use AI effectively in teaching, and among secondary teachers 80 per cent, more than the international averages of 74 and 75 per cent. Almost half of primary teachers feel overwhelmed by the arrival of new technology in education.
How solid is this? The obligation is settled and enforcement has been possible since August 2026, but the track record is short. There are no fines yet, no Dutch enforcement practice and no second measurement of how organisations are handling it. The legal text, the Commission and the Dutch supervisor do already point the same way.
For the leadership team
Can the organisation show today which AI systems are in use and what the people working with them have learned in order to do so? That is the question a supervisor asks after an incident. In most organisations nobody yet owns the answer, while seven in ten employees say the demands on their skills have already shifted.
Sources
- CuatrecasasDigital Omnibus on AI has been published
- Boston Consulting Group · Vinciane Beauchene, Sylvain Duranton, David Martin, Vanessa Lyon and Jeff WaltersAI at Work: Why Strategy Matters More Than Tools
- Autoriteit PersoonsgegevensVerder bouwen aan AI-geletterdheid (tweede handreiking)
Tegelijkertijd is er geen verplichting om het niveau per individu te meten, of AI-geletterdheid certificeringen te behalen.
- Kohnstamm Instituut, IVA Onderwijs, UvA, Fontys and Ipsos I&O · Marieke Buisman et al.TALIS 2024: De werk- en leeromgeving op school in beeld. Nederlandse resultaten van de internationale TALIS-survey
Maar liefst 80% van de Nederlandse leraren geeft aan onvoldoende kennis en vaardigheden te hebben om AI toe te passen in hun onderwijs.
- European CommissionAI Literacy: Questions & Answers
The supervision and enforcement rules apply from 2 August 2026 onwards.
- EU Artificial Intelligence Act (artificialintelligenceact.eu)Artikel 4: AI-geletterdheid (AI-verordening)
We summarise in our own words and quote no more than one sentence per source. Read the original: there is more in it. How we use sources.
Academic research
- Technische Universität Darmstadt · Benlian et al.Scaling AI literacy in organizations: A framework and roadmap for managers
- Rheinland-Pfälzische Technische Universität Kaiserslautern-Landau · Krenkel and AurichDeveloping AI Literacy in Manufacturing: A Role-Specific Training Framework for Lean Production Systems
Academic articles on the same phenomenon, no more than a year old at the last search, from universities in Europe and North America, found through the open database OpenAlex. These articles count towards the publications per month; the text and key figures above are based on the sources. How we choose the articles.
How this briefing has been updated
Title now describes the topics; previously “AI literacy is the law, and since August enforceable too” (AI)
Added 2 articles from universities in Europe and North America (via OpenAlex). (AI)
First draft based on the legal text, the European Commission's Q&A, the guidance from the Dutch Data Protection Authority, Cuatrecasas and BCG. (AI)