Good question. This just comes down to our definition of blocking. Condition 1 is more intuitive so let's start there, if we have a chain (A->B->C) or a fork (A<-B->C), B is said to block the path from A to C.

However if we have a collider (A->B<-C) the path from A to C is "already blocked". And if we are to consider B in our blocking set it would not satisfy Condition 2.

A way to understand this is blocking is just a statement of conditional independence. In the chain and fork above, conditioning on B makes A and C statistically independent. However in the collider, A and C are already statistically independent. And if one does condition on B, we actually create a spurious association between A and C. This is also known as Berkson's Paradox.