The European Business Community takes note of the ambition to reach a General Approach on the AI Act at the Transport, Telecommunications and Energy Council meeting on December 6th 2022. While good progress was made in the past six months, we believe more could have been done if time allowed.
We particularly want to commend the Council’s decisions on some challenging questions and provisions, especially around the definition of Artificial Intelligence and the classification of highrisk AI Systems.
Yet despite this progress we are still gravely troubled that the overall direction of the AI Act will significantly limit Europe’s AI uptake and utilisation. At a moment where Europe must be geopolitically strong, we cannot afford to lose in the AI race.
As a result of these concerns, we bring to attention some of the outstanding issues that remain in the potential General Approach that will be imperative the Council clarifies and defends during negations with the co-legislator.
• Aligning the Definition of AI- The Council’s efforts to give a focused definition of AI systems is highly positive, reflecting the idea of autonomy in the definition helps align with the work done at the OECD level1 . The Council must continue defending this approach in the future negotiations.
• High-Risk Classification Needs More Nuance- We congratulate the Council for finding a compromise that seeks to avoid blanket classification of high-risk systems. However, we note the Council position still leaves a wide scope for high-risk categories, overlapping and capturing many areas which are already comprehensively regulated by existing product safety laws. In addition, Annex III provides too broad of interpretation and no clear methodology for defining a high-risk system. We are concerned that leaving such crucial issue up to an implementing act will not provide investment certainty in the shortterm as the AI ecosystem waits for the implementing act. The Council should seek to find a compromise with the European Parliament that avoids prolonging uncertainty with secondary legislation but defends the spirit of having an unequivocal way of identifying high-risk usage.
• General Purpose AI- The broad regulation of General Purpose AI Systems (GPAI) defies the logic of the risk-based approach of the AI Act as well as the Commissions’ own drafting which did not include GPAI. This will cause Europe to be cut off from systems that utilise GPAI in high-risk areas and will result in a differentiated market for various AI products. Additionally, the knock-on effect on open source is unknown. In light of this it is crucial that an impact assessment be carried out in full consultation with all stakeholders, which addresses the fair and proportionate allocation of responsibilities and distribution of information between the various actors of GPAI in the supply chain.
