Every organisation that deploys or uses AI systems in the EU must ensure that their personnel have a sufficient level of AI literacy. This applies to every employee, in every sector, across all 27 member states.
Feb 2025
Already in force
27
EU Member States
All
Sectors Covered
7.5M+
Max Fine (EUR)
Already in force. Article 4 became binding on 2 February 2025 as part of the first phase of the EU AI Act. Organisations that have not yet implemented AI literacy measures are in potential non-compliance.
“Providers and deployers of AI systems shall take measures to ensure, to their best extent, a sufficient level 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 in which the AI systems are to be used, and considering the persons or groups of persons on whom the AI systems are to be used.”
— Article 4, Regulation (EU) 2024/1689 (EU AI Act)
Captures both organisations that develop AI systems and organisations that use them. A law firm using AI legal research, a bank using AI credit scoring, a hospital using AI diagnostics — all are deployers.
Extends beyond employees to contractors, consultants, temporary workers, and anyone else who interacts with AI systems on the organisation's behalf. Not just the IT department.
A proportional standard — the required level depends on the person's role, the system's complexity, and the context of use. All personnel need some level; none can be excluded.
Provides proportionality but does not eliminate the obligation. Organisations must make genuine, documented efforts. Ignoring the requirement because training is expensive is not compliant.
Article 4 does not allow organisations to train their technology staff and consider the obligation met. Every function that touches AI — which in 2026 is every function — must be addressed.
AI Tools Used
Applicant tracking, resume screening, sentiment analysis, workforce planning
Literacy Required
Understanding bias in AI screening, when to override algorithmic recommendations
AI Tools Used
Content generation, customer segmentation, predictive analytics, personalisation
Literacy Required
Hallucination risks in AI-generated claims, ethical AI-driven targeting
AI Tools Used
Fraud detection, financial forecasting, automated reporting, compliance monitoring
Literacy Required
Reliability limits of AI predictions, automation bias, audit implications
AI Tools Used
Demand forecasting, inventory optimisation, quality control, logistics
Literacy Required
Model prediction limitations, training data divergence, human oversight
AI Tools Used
Contract review, legal research, regulatory monitoring, compliance assessment
Literacy Required
Functional AI literacy for own work plus governance literacy for Article 4 advisory
AI Tools Used
Strategic AI investment, risk oversight, deployment governance
Literacy Required
Strategic AI understanding, regulatory alignment, ethical AI oversight
or 1% of worldwide annual turnover
Minimum penalty tier for general violations including Article 4 non-compliance.
Risk multiplier for other violations
AI literacy failures may aggravate penalties for high-risk system violations. The literacy obligation is foundational.
National authorities enforce
Each member state designates enforcement authorities. Approach varies by jurisdiction but the obligation is uniform.
Beyond regulatory fines
Organisations whose personnel misuse AI due to insufficient training may face civil liability, client claims, and reputational damage that exceeds any regulatory fine. The risk is operational, not just compliance.
Based on the text of Article 4 and emerging guidance from national authorities and legal scholars, compliance requires these elements.
Map every AI system your organisation uses and identify who interacts with each one. Many organisations will be surprised by how pervasive AI is in their existing technology stack.
Create a matrix mapping organisational roles to the AI systems they interact with and the level of literacy each interaction requires. Cover all functions, not just technology and legal teams.
Implement role-appropriate training calibrated to the needs assessment. The receptionist using AI-assisted scheduling needs different training than the data analyst building predictive models.
Record who was trained, when, on what content, and to what standard. Article 4's "to their best extent" language requires demonstrable compliance efforts.
Verify that training actually develops the required competence. Simply providing training is insufficient if it does not produce sufficient understanding.
AI literacy is not a one-time event. Update training programmes as AI capabilities, organisational use, and regulatory guidance evolve.
4. panta MI pratibas prasibai ir konkretas, tuliTejas sekas tam, ka juridiskie specialisti strada ar MI rikiem.
Darmstates Apgabaltiesas spriedums Vacija radija specIgu precedentu: kad tiesa nozimets medicinas eksperts plasi izmantoja MI bez atklasanas, tiesa noteica eksperta honoraru nulles eiro apmera un pasludinaja visu zinojumu par nepielauJamu.
MI riki tiesvediba tagad palidz ar judikaturos izpeti, dokumentu parskatisanu un prognozejoso analitiku. Juristiem jasaprot, ka generativais MI var radit neesosas lietu atsauces.
Ligumu sagatavovsana, pienaciga parbaude un regulativa atbilstiba arvien vairak ietver MI palidzibu. MI genereta liguma valoda prasa cilveka parskatisanu.
Intelektuala ipasuma praktikiem, kas izmanto MI precu zimju meklesanai vai patentu analizei, ir nepieciesama specializeta pratiba, tostarp MI ierobezojumu atzisana novitates novertesana.
Juristiem, kas konsulte regulativajos jautajumos, nepieciesama MI riska klasifikacijas sistemu izpratne, zinasanas par nozarei specifiskiem regulejumiem un kompetence MI ligumisku noteikumu joma.
Pirma ES dalibvalsts, kas pienema visaptverosa valsts MI likumdosanu. Obligata klienta informesana par MI izmantosanu juridiskaja parstaviba.
Darmstates Apgabaltiesas spriedums noteica obligatu MI atklasanu visiem ar tiesu saistitiem iesniegumiem.
Latvija, Lietuva un Igaunija koordine parrobezu MI regulejuma ieviesanu juridiskajiem pakalpojumiem.
The comparison to GDPR is instructive. When GDPR took effect in 2018, it created massive demand for data protection training. The EU AI Act is likely to follow the same pattern.
EU AI Act: Regulation (EU) 2024/1689 - Full Text
https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32024R1689
European Commission: Regulatory Framework for AI
https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
AI Act Article 4: AI Literacy - Analysis and Commentary
https://artificialintelligenceact.eu/article/4/
NOVA School of Law: AI Regulation Research
https://novalaw.unl.pt/en/ai-regulation/
European Parliament: EU AI Act - First Regulation on Artificial Intelligence
https://www.europarl.europa.eu/topics/en/article/20230601STO93804/eu-ai-act-first-regulation-on-artificial-intelligence
Future of Life Institute: EU AI Act Explorer
https://artificialintelligenceact.eu/
