Monopolisation or centralisation has been a long-standing concern in AI policy. Access to data confers a first-mover advantage on a small number of companies; the resources required to train large AI models give those same companies a competitive advantage in a field increasingly interested in foundation models; and access to knowledge and skills is unevenly distributed. History suggests that disrupting the monopolies surrounding general-purpose technologies requires policy intervention, and that such intervention typically takes decades to deliver results. Investments in European AI to grow the AI research community and connect that community to local innovation ecosystems – sharing the benefits of AI across the continent – can help disrupt this influence. Regulatory interventions to disrupt the power asymmetries that shape AI development also play a role.