Anthropic Claude Mythos Preview: The More Capable AI Becomes, the More Security It Needs | AI Sec Watch