AI Act: AI use only with sufficient competence
On February 2, 2025, a new stage of the AI Regulation will come into force. There are bans and a competence obligation for all employees.
(Image: Shutterstock/Sansoen Saengsakaorat)
AI systems with unacceptable risks are now prohibited. This applies to social scoring applications, for example, which monitor citizens. Not that these systems have ever existed in the EU before. However, the AI Act, which came into force last August, ensures that such systems will not be used in the future. The ban will apply from February 2, 2025, as the law will be applied in stages.
In addition to the ban on some AI applications, the date also entails other obligations. Among other things, employees may only use AI in the workplace if they are sufficiently competent to do so. What exactly this means, however, is somewhat vague. "Providers and operators of AI systems shall take steps to ensure, to the best of their knowledge and belief, that their personnel and other persons involved in the operation and use of AI systems on their behalf have sufficient AI competence", states Article 4 of the AI Act. Competence therefore means "technical knowledge, experience, education and training", and users must also understand the context in which they wish to use AI.
This obligation applies to all companies, regardless of size or focus, and also to all AI applications, regardless of the risk level to which they are categorized. What is not entirely clear is who is considered an operator. It is questionable whether employees who use ChatGPT in their day-to-day work, for example, are affected, or whether it is a matter of having set up your own chatbot – and thereby operating an AI application.
AI regulation for general purpose AI
The AI Act itself does not initially provide for penalties if employees do not have sufficient competence. Determining this is also likely to be quite difficult – but it is probably still possible. In principle, the penalties set out in Article 99 are not envisaged until August 2, 2025. Until then, the individual member states must also first appoint national authorities to monitor the AI Act.
In three months' time, a guideline will also be published for general purpose AI (GPAI) models, which will describe how to use them in a legally compliant manner. A further three months later, i.e. twelve months after the AI Regulation comes into force, the obligations for GPAI will also apply, just like the penalties. These include a technical documentation obligation, the as yet unresolved issue of copyright and AI, the disclosure of training data and more.
(emw)
