Your "concentration of power" is just two labs making models that most people prefer the last couple of months. Neither has more access to capital and resources than Google, more ability to pivot quickly than Xai, more access to labor than all of the Chinese labs, etc. How do you keep from a "concentration of power" without just forcing subsets of the population to use a known lesser model, or purposely kneecapping Research and Development at the labs that currently have the best models?