AI firms will race to build dangerous AI even if it costs them money

Competition among AI companies makes them prioritize speed over safety, even when it's not in their best interest. This means the market structure itself, not just technical choices, drives how risky advanced AI becomes.

Everyone assumes competition makes things better and safer. This paper shows that for advanced AI, competition might make it more dangerous. Companies might build AI that poses existential risks, not because they want to, but because the market forces them to race. This means governments might need to intervene in the market structure of AI development, not just regulate the technology itself.