Building AI Literacy under the AI Act: Best Practices

On 1 August 2024, the European Artificial Intelligence Act (AI Act) entered into force, aiming to ensure that AI developed and used in the EU market is trustworthy, lawful, ethical, and robust. From the 2nd of February 2025, the first two key provisions of the EU AI Act come into full effect, namely Article 4 (AI literacy) and Article 5 (prohibited AI practices). These are especially relevant for providers and deployers under the AI Act. In this blog article, we explore what Article 4 encompasses and how your organization can comply with its requirements.

‍

What is Article 4?

Article 4 of the EU AI Act mandates that providers and deployers of AI systems must take measures to ensure a sufficient level of AI literacy among their staff and other individuals involved in the operation and use of AI systems. The level of literacy required should take into account employees' technical knowledge, experience, education, and training, as well as the context in which AI systems are deployed.

“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.”
- Article 4 of the EU AI Act

AI literacy, as defined in the AI Act, is a combination of knowledge, understanding, skills, and experience. Organizations must ensure that their employees:

• Understand AI fundamentals and their role in AI governance.
• Can interpret AI-generated outputs correctly and recognize potential biases or risks.
• Can make informed AI deployment decisions.
• Can navigate compliance requirements under the EU AI Act.

Who is affected?

In general, the EU AI Act lays down two main actors in the AI value chain who are affected by the provisions, namely

• Providers of AI systems:  this includes natural or legal persons, public authorities or agencies, and any other entity that can develop and AI system or a GPAI model, or that has an AI system or GPAI model developed, placed on the markets, or integrated into their system and service under their own name or trademark.
• Deployers of AI systems: this includes natural or legal persons, public authorities or agencies, and any other entity that is using an AI system in an organization (which excludes the use for personal, non-professional activity).

Article 4 specifies who is obliged to be AI literate, namely anyone dealing directly with AI systems, which includes staff of providers and deployers, employees working on their behalf, as well as contractors or service providers. As the use of Generative AI in organizations increases, both for own use in the professional context, as well as integrated into systems or services, organizations have a high chance to be affected by the provisions of Article 4.

But additionally, it is recommended to foster a general understanding of AI within the company regardless of the maturity of AI, as employees might be unaware of the AI Act's provisions and the scope and risks of AI systems they might already use for work even without the company's approval. Providing an AI literacy program covering the fundamentals of AI helps your employees to stay innovative while preventing and preparing for potential risks.

Why is it Important to Ensure AI Literacy?

AI literacy is a fundamental pillar of effective AI governance, which in turn plays a crucial role in maintaining a competitive edge through responsible innovation and enabling the adoption of emerging technologies. However, the importance of fostering AI literacy extends beyond these strategic advantages. There are several additional reasons why investing in AI education and awareness from the outset is essential:

1. Regulatory Compliance – Avoiding non-compliance risks with EU AI Act regulations.
2. Responsible AI Usage – Ensuring AI is used ethically and effectively.
3. Risk Mitigation – Reducing errors and biases by training employees to properly interact with AI.
4. Trust and Transparency – Building confidence among customers, stakeholders, and regulatory authorities, which builds reputation in the long run.

Translated into practice this means that companies should…

• Provide information and training on at least the basics of AI, and more in-depth training depending on the role and context,
• Provide information about each AI system in use, and how it might affect people or target groups,
• Inform about the do’s and dont’s of each used AI system in the company,
• Give specific guidelines on how to interpret the output, and specifically, how to detect inaccurate outputs, and what to do in such a case,
• Explain the basics of the EU AI Act, especially when interacting with a high-risk AI system.

What Happens if an Organization is not Compliant?

While Article 4 compliance is not strictly enforced, organizations can face serious consequences, both monetary and reputational, if they’re not compliant. In the event of a compliance issue, authorities may evaluate whether employees have received adequate training, and failure to demonstrate adequate AI literacy initiatives could lead to regulatory scrutiny and reputational harm. Moreover, companies could face fines if an employee, due to gross negligence, causes an AI-related incident or lacks the necessary compliance knowledge, resulting in a regulatory violation. Additionally, liability risks arise if damages occur due to insufficient AI training. Regarding the sufficiency of training endeavors, the AI Office has made it clear that merely providing written instructions on AI usage is inadequate. Article 4 explicitly requires organizations to offer structured guidance and training to ensure employees can operate AI systems responsibly and in compliance with regulations, according to the European AI Office.

‍

Translating Article 4 into Practice

A potential roadmap for the successful implementation of AI literacy could look like the following…

1. Determine whether you need to comply

Determine whether you are a developer or deployer of an AI system, for instance by taking our free EU AI Act Compliance Checker.

2. Identify the levels of AI literacy needed among your workforce depending on their roles and context of AI systems

In each company, the following groups of workforce are most likely:

• General workforce and operational teams: Employees use AI systems in a professional context to increase productivity or efficiency for tasks.
• Developers: Employees developing and monitoring AI systems and outputs, and who actively mitigate risks. Involved in the AI governance process.
• Risk & Compliance: Employees are actively involved in risk management and governance and need in-depth regulatory knowledge.

Depending on their role and the context of the AI system(s) in use, the requirements for levels of AI literacy can vary.

3. Find the right training or program

1. Create your own training: Leverage your existing educational department and develop your own training, covering the aforementioned knowledge and skills.
2. Find a new training for your needs on our AI Training Marketplace: Browse existing trainings and select a program or course depending on your needs and budget.
3. Use trails AI fundamentals training: For your general workforce, the trail AI governance platform comes with an already integrated "AI Fundamentals" training.
4. Create trainings within trail or integrate existing AI-specific trainings from your Learning Management System into trail, to further centralize AI compliance management and evidence tracking.

4. Implement it in your organization and update regularly

To ensure holistic AI governance within your organization you need to assign responsibilities for implementation and monitoring of the trainings. Moreover, documenting these steps can be helpful for potential audits and compliance.

Do you want to get a roadmap for successfully implementing AI trainings in your organization? Download our step-by-step checklist here!

‍

AI Literacy Best Practices

The European AI Office has released a living repository of best practices to support the implementation of Article 4 under the EU AI Act. It contains data from surveyed companies, who signed the AI Pact and are sharing their implementation best practices. Here is a summary of the AI literacy best practices from European organizations:

• Most organizations use a structured e-learning as approach and embed the trainings in an on-demand, self-paced online platform,
  • but also workshops and in-person sessions,
  • or video and podcast trainings are used,
  • and accessibility for all employees is ensured, for instance, by adding audio-descriptions or subtitles.
• Most organizations also provide role-specific training, dividing the AI literacy program into one part focused on the “general understanding” of AI and one part focussed on  role-specific and organization-specific training for single use cases, contexts and industries.
  • Most also provide different training based on the level of experience: from beginner/basic to advanced, tailoring the training not only to the need of the role and the context of the AI system, but more specifically to the current state of knowledge of each user, employee or affected user.
• Some embed it in a structured “upskilling program” that takes into account the ever changing roles and responsibilities that come along with emerging technology (for instance, Assicurazioni Generali S.p.a. established a “New Roles School” in collaboration with universities to develop specialized AI roles; see Living Repository of AI Literacy Practices - v. 31.01.2025, p. 4).
• Some are leveraging game-based and story-driven learning methods;
  • This even includes the use of virtual reality simulation to provide hands-on experience with AI tools.
• Some are launching employee-led communities to promote an AI culture via blogs, webinars and interviews and fostering sharing AI expertise across departments and functions.
• If the training is done in-house, the development of the training content is done by or in alignment with employees developing the AI systems or have a deep understanding of the technology. Those leading AI governance within the organization should also participate in the coordination process.
• In more mature companies, AI training is already integrated into the onboarding process.
• One company developed an “AI Governance Champion” or "Responsible AI Champion" program, which serves as an intermediary between business units and AI governance frameworks.

In summary, the current best practices balance the tailoring of courses and their scalability, enabling the general workforce to responsibly use AI on a day-to-day basis, while equipping more advanced roles also with more advanced skills and knowledge. Moreover, whether and whatever Learnings Management System an organization has, the AI literacy training is integrated into clear structures, including a platform that hosts the training and a review mechanism. In general, trainings and programs are developed in close collaboration with the respective intended roles and context, both of the AI system in use, but also the organization as a whole. This means that a training program for employees of an insurance company will vary regarding content and focus when compared to a health care provider or a design studio, which uses AI-powered design features. A one-size-fits-all solution is therefore only appropriate in a very general, fundamental and basic context of AI systems, for instance, the professional use of conversational or generative AI for daily tasks. As soon as there are more specific AI systems in place, when AI is developed in-house or part of a product or service, a tailored training seems to be the most appropriate practice. This tailoring can and should be done in alignment and close collaboration with the employees working on or with AI to build a well-working AI literacy program.

‍

Conclusion

Article 4 of the EU AI Act mandates that providers and deployers of AI systems must take measures to ensure a sufficient level of AI literacy among their staff and other individuals involved in the operation and use of AI systems. With the rapid adoption of enterprise AI and generative AI models, it is likely that many organizations fall under the scope of Article 4 and, hence, must take steps to comply now. There are already many best practices that show how successful implementation can look like and how to balance generic, fundamental programs with role-specific trainings.

When searching for the right external provider of AI trainings, you can browse this AI training marketplace and find a program that fits your needs and budget.

If you want to kickstart your AI governance, leveraging a platform or GRC-tool can facilitate your governance processes - be it for procured or internally developed AI use cases. trail is putting holistic AI governance intro practice right from the start, by not only integrating the management of all your AI trainings directly in trail, but also collecting all your AI systems in one central place. Our platform is also helping you to provide more technical evidence to prepare for audits and regulatory compliance. Check out how trail’s features can transform your AI governance journey here!

‍