Further provisions of the European AI Act will come into force on 2 February. These include bans on certain AI practices such as social scoring systems, manipulative AI techniques or emotion recognition in the workplace. In addition, requirements for AI skills for employees will come into force. Susanne Dehmel, member of the Bitkom management board, explains:
‘The AI Act should ensure legal certainty for artificial intelligence in Europe – currently, the exact opposite is looming. When further provisions of the European AI Regulation come into force from Sunday, it is unclear for which applications the legal ban will actually apply. Politicians have set high standards and tight deadlines for companies in the AI Act, but have not done their homework themselves. The risk is borne by the companies that develop or use AI. While the US is planning to spend hundreds of billions of dollars on AI development and China is publishing extremely powerful language models, we in Germany and Europe are putting obstacles in the way of AI companies.
It is by no means the case that the definition of prohibited practices only covers clearly problematic applications. Only today, two days before the regulation comes into force, does the EU Commission intend to publish urgently needed guidelines on the definition of AI and prohibited practices. Companies whose systems fall into the numerous difficult-to-classify borderline cases will find out today, Friday, whether their affected systems may remain on the market or must be taken off the market over the weekend.
There are also still uncertainties regarding the so-called AI competence requirements, which affect practically all companies developing or using AI. All companies that use AI must ensure that the employees concerned have ‘a sufficient level of AI competence’. It is still unclear when this obligation will be met because no corresponding supervisory authority has yet been set up in Germany that could provide guidance to businesses. There is also a lack of specific information from the European side. However, companies should not let this deter them from continuing their AI efforts, but rather use the deadline as an opportunity to train their employees in how to use AI and thus make them fit beyond the AI Act.
We need more German and European AI. To achieve this, we need – in addition to money and more support for European AI companies – a regulatory framework that is truly innovation-friendly, not an excess of vague regulations, but instead a minimum of binding and clear guidelines for practical application. It is high time to start seeing artificial intelligence not primarily as a threat, but as an opportunity.’