The article doesn't establish whether AI ultimately benefits attackers more than defenders. Open weight models give defenders access to the same capabilities. At the very least the community has a level of symmetry they wouldn't otherwise have.