Positive knots have negative signature

It was asked by J.Birman, Williams, and L.Rudolph whether nontrivial Lorentz knots have always positive signature. Lorentz knots are examples of positive braids (in our convention they have all crossings negative so they are negative links). It was shown by L.Rudolph that positive braids have positive signature (if they represent nontrivial links). K.Murasugi has shown that nontrivial, alternating, positive links have negative signature. Here we prove in general the old folklore conjecture that nontrivial positive links have negative signature.