This is s misunderstandig of Hoares “billion dollar mistake”. The mistake was not the existence of nulls, the mistake was type systems where every reference could be null.
SQL absolutely does have that problem. You have no way of enforcing or checking that an expression is supposed to be non-NULL; at best you get an error at runtime when you try to insert it into a given column, which is usually far too late.
No, it's a specific problem of SQL that you can apply constraints to stored or quasistored columns but not to expressions. In a real typed language every expression has a type.
