How to understand the AI Literacy?

Pursuant to Article 3(56) of the AI Regulation:  AI Literacy means skills, knowledge and understanding that allow providers, deployers and affected persons, taking into account their respective rights and obligations in the context of this Regulation, to make an informed deployment of AI systems, as well as to gain awareness about the opportunities and risks of AI and possible harm it can cause.

Although the AI Regulation does not indicate any specific types of competences, it does define three basic categories of such competences that should be developed by Providers and Deployers of AI Systems. These are competencies in the area of:

• skills (i.e. expanding practical knowledge of AI and applying this knowledge in everyday practice);
• knowledge (i.e. expanding theoretical knowledge of AI);
• understanding (i.e. the ability to understand how AI works and how it affects the processes in which AI is used, e.g. the ability to apply the knowledge in practice in non-standard or non-obvious situations where the AI system will use false data or will be guided by prejudices).

All three of the abovementioned areas should, in principle, be included in the development and training goals of the organization.

Setting goals regarding AI skills based on risk levels and roles

The AI Regulation also does not contain any specific regulations that would define the guidelines or framework that AI training would have to meet, allowing for a great deal of discretion in this regard as to what is important for the organisation and its employees.

In any case, however, the aforementioned training is intended to ensure that all persons who have, to a greater or lesser extent, contact with AI acquire comprehensive knowledge and skills, as well as understand the tools they interact with and the risks involved. In other words, AI Literacy is not just about the technical skills of handling AI systems, but also about responsible assessment of the risks and potential damage these systems may cause to an organisation. Equally important, AI Literacy responsibilities do not end with a one-off training, but also include regular ‘retraining’ of employees to ensure that the knowledge and skills they have acquired in the field of AI systems remain up to date, i.e. are maintained at a level that meets the condition of the employee having an appropriate level of AI Literacy.

In practice, it is pointed out that the obligation to ensure an appropriate level of AI Literacy does not apply to all employees working in a given organisation, but only to those employees (or entire departments) who use AI systems in their work – directly or indirectly, to a greater or lesser extent. What is more, such training must be, in each individual case, tailored to the actual needs and scope of duties of a given employee, so that an employee who has little contact with AI in his or her work does not need to be trained as extensively as an employee whose work makes greater use of AI systems. In other words, not every employee who uses AI systems in his or her work is to be an AI expert, and AI training policies need to be ‘tailor-made’ for both the organisation and the employee. In summary, the contextual reference to the appropriate function of the employee must be taken into account. This makes sense because the situation in marketing is quite different to that in HR, for example.

It should also be made clear to employees, as part of their AI training, what other responsibilities they have when using AI. This includes not only the obligations under the AI Regulation itself, in particular raising awareness of high-risk AI applications as referred to in Article 6 of the AI Regulation, or the transparency obligations under Article 50 of the AI Regulation, but also the obligations arising from legal acts such as:

• any copyright law,
• any data protection law, including the GDPR.

As an aid, it is possible to rely on the arrangements that were included in the so-called AI Pact, i.e. an agreement signed with the entry into force of the AI Regulation by more than 100 international companies (both SME sector and large corporations) [hereinafter: “AI Pact”]. The parties to the AI Pact have committed to voluntarily promote the trustworthy and safe development of AI systems. In connection with the aforementioned agreement, the European AI Office has developed a repository that collects and describes some of the current practices used by the signatories of the aforementioned agreement, in order to provide examples of how AI Literacy obligations can be implemented and fulfilled.

Key steps in building AI Literacy in a company

There are four key steps in building AI Literacy in a company:

1. Identify how AI is currently being used and assess the risks involved.
2. Identify key stakeholders and define the goals of AI Literacy based on roles and responsibilities.
3. Develop a structured training program and keep it updated, taking into account the new areas of AI in the organisation and both the technological changes and the changes in applicable regulations.

Incorporate AI skills into ongoing employee development initiatives with clear management and oversight.