EU AI Act Article 4: The AI Literacy Mandate

What is Article 4?

‍

Article 4 is the last article in the first chapter of the AI Act outlying general provisions. It states that:

‍

“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 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 provision is one of the first mandates of the EU AI Act that came into effect on the 2^nd of February 2025. Therefore, time is running out to implement necessary changes and allocate appropriate resources for compliance with Article 4.

‍

However, at first glance, the AI literacy mandate appears broad and somewhat vague, VERB phrases like “to their best extent” and “a sufficient level.” Although Recital 20 offers additional clarification, both still raise many questions. What exactly is required here, and from whom? What are the consequences for negligence or failure to comply by the date of enforcement?

‍

In this article, we are going to simplify Article 4 and Recital 20 into a practical five-step guide to support your journey to compliance with the EU AI literacy mandate.

‍

What is AI Literacy?

‍

Let's start by defining the core concept of Article 4:

‍

AI literacy is the ability to understand, apply, and assess AI technologies while being able to critically reflect on the ethical implications of them. AI literacy equips individuals and organisations with the necessary skills to interact effectively and consciously with AI, whether in private or professional settings. Closely related to digital literacy, it builds on foundational digital and computational skills to promote informed and responsible AI use.

‍

Why is AI Literacy Important?

• Technical Competence: A basic grasp of AI, including machine learning, enables users to effectively and safely utilise AI tools in both work and daily life, ensuring competitiveness in the job market.  

• Informed Decision-Making: A basic understanding of AI's capabilities, limitations and ethical implications empowers users to make informed decisions about when and how to use AI tools, as well as how to interpret their outputs. For example,  recognising that AI chatbots generate responses based on patterns prevents blind trust and encourages fact-checking.

• Responsible AI Use: AI literacy promotes ethical awareness, responsible use, and accountability by helping users understand AI's impact on issues like bias, fairness, and privacy. It also empowers users to demand transparency and accountability from developers. For example, understanding the technicalities of AI and the ethical discourse around it makes developers and users more conscious about biases embedded in training data and outputs.

In short, widespread AI literacy is key to the responsible and safe development of AI technologies. It plays a crucial role in anticipating and preventing risks, such as misuse and dissemination of AI-generated misinformation, by cultivating technical competence and ethical awareness. As AI continues to shape political and social spheres, thoughtful engagement and responsible use are essential for navigating an AI-driven world.

‍

What does Article 4 mean for my Organisation?

‍

Let us now turn our attention to the practical ramifications of the AI literacy mandate. We will begin by clarifying the parties subject to the literacy mandate, followed by a discussion of the potential implications of non-compliance. Finally, we will outline a five-step guide to help you get started with meeting the requirements of Article 4.

‍

Who does Article 4 Apply to?

‍

Which Operator Class does Article 4 Apply to?

Article 4 makes it clear that the responsibility for maintaining AI literacy falls on both AI system providers and deployers (or users). This means that organisations developing AI systems for the EU market, as well as those integrating AI into their business operations, are required to ensure adequate AI literacy within their teams. You can find out more about the operator classes and their obligations within the EU AI Act on our blog.

‍

Which AI Systems are affected?

Neither Article 4 nor Recital 20 outlines specific AI literacy obligations based on the varying risk levels of AI systems. This suggests that providers and deployers of all AI systems—whether minimal, limited, or high-risk, and including General Purpose AI Systems—share equal responsibility for ensuring AI literacy.

‍

However, the risk level of a system could become relevant in cases of regulatory and transparency violations, where negligence in AI literacy may serve as an exacerbating factor for fines and penalties. Furthermore, considering that the EU AI Act is grounded in a risk-based approach, the AI literacy mandate may be understood within the framework of gradual responsibility—the higher the risk, the greater the responsibility to ensure adequate AI literacy.

‍

Who should be AI literate?

Article 4 specifies that “Staff and other persons dealing with the operation and use of AI systems on their behalf” should receive training and education appropriate to their context and role.

• Staff: It is important to ensure that AI literacy is not limited to the IT department but is integrated throughout the entire organisation. While not every employee needs expert-level training, it is crucial that everyone has a basic understanding of AI, its ethical implications, and its specific applications within the company, especially given that 32% of Europeans still lack basic digital skills, according to the 2023 EU report on the digital decade. In addition to this foundational knowledge, role-specific training should be provided to ensure that those working directly with AI outputs or tools are properly equipped.

• Others: The term “other persons dealing with the operation and use” can encompass a wide range of stakeholders specific to each organisation. When developing an AI literacy plan, companies may need to consider importers, distributors, external staff, or even private users, tailoring training and information to suit how each group interacts with the AI system. For example, deployers may require further training and guidance from the AI provider to fully meet their own AI literacy obligations.

In short, all relevant stakeholders in the AI value chain must have the necessary knowledge to comply with company policies and regulations, as well as to make informed decisions when developing, managing, or operating an AI system and its outputs.

‍

What are the Consequences for Non-Compliance?

‍

At the time of this article's publication, the European Commission has not yet defined what constitutes a breach of Article 4, nor imposed fines or disciplinary measures for non-compliance. However, this does not mean ignorance will be without consequences.

• Legal Consequences: Although a breach of Article 4 is not officially considered a punishable offence on its own, negligence may still have legal consequences. According to Article 99 (7) of the AI Act, in determining whether to impose a fine and its amount, “(…) all relevant circumstances of the specific situation shall be taken into account (…)” This implies that if an AI system provider fails to comply with any part of the EU AI Act, legislators may consider shortcomings in meeting the AI literacy mandate when determining a fine. Furthermore, this could indicate that the AI literacy mandate could be summative to other AI Act requirements, such as the transparency requirement.

• Reputational Consequences: Prioritising AI literacy both within and outside an organisation reflects a commitment to responsible and transparent AI development. Conversely, neglecting this requirement may risk undermining the trust of stakeholders like investors, employees, and users, with potentially destructive consequences for a company’s reputation and revenue.

• Wider Societal and Industry Impact: Experts and lawmakers agree that AI literacy is a critical step in preventing irresponsible innovation and misuse of AI technologies. Failing to acknowledge the responsibility of fostering widespread AI literacy can significantly impact the overall safety and trustworthiness of the AI sector. Without ethical oversight, unchecked AI development could lead to irreversible harm, including threats to fundamental human rights and democracy.

How do I ensure Compliance with Article 4?

‍

In order to “(…) take measures to ensure, to their best extent, a sufficient level of AI literacy (…)” you can take inspiration from our comprehensive five-step guide to compliance with Article 4.

‍

1. Stay Informed about AI Act Updates

‍

The first step to complying with Article 4 is staying informed on EU publications related to the AI Act. Member states, the Commission, and relevant institutions are expected to develop codes of conduct and offer compliance support. In Recital 20, the European Commission acknowledges its responsibility to promote AI literacy, committing to work with the European Artificial Intelligence Board to “(…) promote AI literacy tools, public awareness, and understanding of the benefits, risks, safeguards, rights, and obligations related to AI systems.”

‍

Therefore, staying informed not only ensures compliance with the latest amendments to the AI Act but also provides an opportunity to leverage approved codes of conducts, educational materials, and tools.

‍

2. Identify Context and Affected Stakeholders

‍

In accordance with Article 4 and Recital 20, AI literacy should be maintained in the appropriate context and tailored to the individuals affected. Therefore, the second step is to analyse the specific context of your organisation and AI system by asking some of the following questions:

• What are the AI system’s intended and unintended functions?

• Where will it be deployed?

• Who is going to operate it within and beyond my organisation?

• Who will be affected by its intended and unintended outputs and how? (This is also called “consequence scanning”).

It is important to note that providers have additional responsibilities to ensure that the technical details and ethical considerations of their AI systems are clearly communicated and understood by its users.

‍

3. Resource Allocation

‍

Once the specific needs related to the context of your AI system and organisational structure have been identified, the third step is to allocate appropriate resources to implement and develop training and educational programmes to ensure that all employees are AI-literate. This may involve allocating a budget for AI literacy initiatives, assigning employees to an AI literacy management team, or hiring additional staff for that purpose.

The wording used in Article 4 and Recital 20, including phrases such as "to their best extent," creates some ambiguity regarding the precise requirements of this provision. However, the penalties and obligations set out in the EU AI Act are designed to consider the limited resources typically available to start-ups and small and medium-sized enterprises. It may therefore be assumed that the same degree of flexibility is applied to Article 4 when allocating workforce and financial resources to compliance efforts.

‍

4. Education and Training Programmes

‍

After the necessary groundwork has been laid, the fourth step is supplying “sufficient” “measures” to ensure AI literacy. However, what is sufficient? And, what measures could your organisation take?

‍

“Sufficient” AI Literacy

The level of AI literacy required will vary across organisations, and across staff within the organisation. In essence, the level of literacy required is a function of how much say someone has in designing and operating the AI system. The more “power” someone has, the more knowledge is required to fulfill the “literacy” requirement.

‍

So any training or educational programs should cover the fundamental concepts outlined in Article 4 and Recital 20, ensuring that providers, deployers, and all stakeholders are equipped to make informed decisions regarding AI systems. This may include education on:

• Technical foundations of AI

• Proper application of technical elements during the AI system’s development and utilisation

• Best practices for ethical, trustworthy, and responsible AI development and utilisation

• AI risks and dangers

• Appropriate interpretation of the AI system's output

• Understanding the effects of AI-assisted decisions

• Ethical and socio-cultural debates around AI

• Regulations and compliance requirements

To address these subject areas, AI literacy initiatives could include hands-on workshops, practical tutorials, seminars, lectures, and be consolidated further in digital or printed guides. Providing interdepartmental and multi-functional learning opportunities can also assist in achieving a more consistent level of AI literacy on a company-wide basis.

‍

Training and Certification by oxethica

‍

At oxethica, we are dedicated to enabling safe, responsible, and trustworthy innovation of AI. We believe that developing the right competencies, skills, and knowledge is the essential first step toward ensuring the trustworthy use and development of AI. Therefore, we offer training and certification programs to help you build and strengthen your AI literacy.

‍

Training: We offer two online courses in partnership with Saïd Business School at the University of Oxford covering the basics of AI and key discussions surrounding AI ethics, regulation, and compliance.

‍

Certification: We are currently working on providing certifications for AI professionals, AI Ethicists, and AI Auditors ensuring competence in accordance with current regulations and insights.

‍

For more information and to enrol today, visit our Training and Certification landing page.

‍

5. Review and Update Literacy Measures

‍

In order to keep pace with the rapid advancements in AI technologies and the evolving legal landscape, it is essential that organisations regularly review and update their AI literacy programmes accordingly. This ensures that your organisation stays aligned with regulatory requirements and remains current with the developments in AI technologies.

‍

In a Nutshell

‍

What is AI Literacy?

‍

AI literacy is the ability to understand, apply, and assess AI technologies while being able to critically reflect on the ethical implications of them. AI literacy equips individuals and organisations with the necessary skills to interact effectively and consciously with AI, whether in private or professional settings. Closely related to digital literacy, it builds on foundational digital and computational skills to promote informed and responsible AI use.

‍

Why is it important?

‍

Ensuring AI literacy is not just a regulatory requirement under Article 4 of the EU AI Act—it is a fundamental responsibility for organisations deploying and developing AI systems. By fostering AI literacy among staff and stakeholders, businesses can enhance technical competence, promote ethical AI use, and mitigate risks associated with misinformation, bias, and regulatory non-compliance.

‍

While the exact penalties for failing to meet this mandate remain undefined, the potential legal, reputational, and societal consequences underscore the urgency of proactive implementation. As AI continues to shape industries and societies, prioritising AI literacy will be essential for building trust, ensuring compliance, and driving responsible innovation in an AI-driven world.

‍