This problem is quite similar to the GAI problem: the more automation we have, the greater damage an attack can have. The question is whether there is any way to limit the biggest attacks that can happen? I don't think, but I wish there would be a solution.