I think that's wrong as well, you can't achieve "regulatory self-capture" you need lobbying and insiders, and a government to get there, at least from my primative uninformed view.
A deadly failure mode here would be for governments to think the AI labs have this under control.