The current EU approach to AI regulation faces several challenges and limitations that need to be addressed. One of the main issues is to what degree product legislation approach is fit for mitigating more systematic and democratic risks of AI systems and the lack of resources for regulatory and enforcement agencies. Another challenge is the compatibility of the EU approach with the collective bargaining and co-determination models that are prevalent in some member states. How can the EU ensure that workers are adequately involved in the governance of AI systems that affect their work environment and conditions? A third issue is the power relations and liability between the providers and the users of general-purpose AI systems. The EU approach seems to be unclear or reluctant to assign liability to the providers of AI systems, which may create an imbalance of power and responsibility between them and the users or deployers of the systems. Finally, the sustainability and material footprints of AI systems are arguably overlooked in the EU approach, despite it being a potentially important leverage point.