KAN position paper on the EU draft regulation on artificial intelligence

A new legal framework is intended to make the Single Market a model of trustworthy and innovative artificial intelligence (AI) and enhance the EU’s competitiveness internationally. KAN has formulated a position paper on the proposed regulation from the perspective of the occupational safety and health lobby.

On 21 April 2021, the European Commission published a proposal for a regulation laying down harmonized rules on artificial intelligence. The regulation’s purpose is to leverage economic potential and ensure that AI is geared to human beings. A further aim is to create legal certainty on what methods and concepts are deemed to constitute artificial intelligence. Trade barriers arising from discrepancies between national regulations are to be avoided, and space and competition for innovation are to be created without prejudice to European fundamental rights.

AI and highly automated applications are among the key developments of our time. Increasingly, they also affect the area of safety. Besides all the opportunities and possibilities it offers, however, AI also gives rise to risks. Its impact on work processes presents challenges for safety, ergonomics, the human psyche and social policy. Changes in skills requirements and work processes, the emergence of new vocations and issues of ethics, data protection and discrimination are among the reasons why AI is growing in significance in occupational safety and health circles. Risk assessment also presents a particular challenge: the capabilities and limitations of complex high-risk AI systems will almost certainly lie beyond the comprehension of human beings tasked with supervising them.

The Commission’s proposal ascribes a key role to standardization (based on the New Legislative Framework) in application of the regulation. The first German Standardization Road­map AI, now being revised by DIN, also emphasizes that requirements concerning AI systems such as transparency, robustness and reliability must be specified at the technical level by harmonized European standards, confidence in AI enhanced and innovations promoted.

It is in KAN’s interests for the regulation to set out suitable and coherent legal requirements and give rise to corresponding standardization mandates. Standards are to be used as a tool for identifying and reducing risks. As the voice of the German occupational safety and health lobby, KAN published its position paper on the proposal for an EU regulation on artificial intelligence on 1 March 2022. This paper is the result of two exploratory discussions between the German Federal Government, the regional governments, the social partners and the accident insurance institutions.

KAN is of the opinion that the following points in particular require clarification:

• Legal basis of the regulation: the proposal contains obligations directly addressed at the persons, businesses and establishments applying AI systems. It formulates, for example, monitoring and information obligations regarding the use of high-risk systems, particularly in Article 29. It is to be clarified to what extent the legal basis of the regulation is sufficient for such obligations, which also include the systems’ operation, and what consequences this has for the users.
• Review of other Single Market legislation with respect to its interfaces with AI: must legislation such as the Low-voltage Directive be amended to take account of the application of AI, in order to ensure a uniform body of legislation?
• Requirements concerning high-risk AI systems: the position paper proposes detailed requirements concerning high-risk systems, for example with respect to supervision by human beings. The mode of operation of more complex AI systems will almost certainly lie beyond the comprehension of persons supervising them. These persons should instead be made aware of the capabilities and limitations of these high-risk systems.

The proposed regulation is currently being negotiated in the responsible committees of the European Parliament. Particular topics of discussion are the definition of AI upon which the regulation is based and the question of what legal arrangement is suitable for the complex subject of regulation to be appropriately addressed. The plenary vote is expected towards the end of the year. In line with the ordinary legislative procedure, negotiations between the Member States have also begun in the Council. Owing in particular to the interwoven ethical, legal, socio-political, technological and economic aspects, it remains unclear when conclusion of the negotiations on the AI Regulation will be possible.

Katharina Schulte
schulte@kan.de