There’s nothing inherent to the protocol that prevents spamming the global namespace; anyone can publish anything by definition. However, there is also nothing that would cause the the spam to be pinned (seeded) or cached if no other nodes request it, and it would not be reachable from any links.

The replication semantics (through Filecoin, extensions to bitswap, ipfs-cluster, or application layer logic like tit-for-tat pinning between protocol participants) will need to account for this, but they are not firmly established yet.