The one model that I think is useful for evaluating the trust required is
Where would proof-of-stake fall in there? Is there difference between different implementations?