What are the most common barriers that prevent AI labs from binding to deliberative outcomes? Which barriers are structural versus contingent on political will?

Ability to make outputs enforceable through known and viable legal structures and mechanisms (e.g. regulation that requires authority to implement or seriously consider the outputs).