This is clearly the exit strategy here. That's why AI companies' try to label their products as "dangerous" and call for their own regulation. On the one hand, they want to be regarded as a matter of "national security" because they bloody well know that this gives them access to taxpayer money, which in turn can bail them out of their failing bet on data center build-out. On the other hand, it helps them shift liability onto regulators if things go south and allows them to say: "We told you we wanted regulation, but you wouldn't listen."