Is there a λProlog that wouldn't need a type system?

Question

I suspect that λProlog needs a type system to make their higher
order unification sound. Otherwise through self application some
Russell type anomalies can appea