Former national cyber director: Anthropic’s ‘Mythos’ AI can hack nearly anything and we aren’t ready

Anthropic's most powerful model yet is less a product launch than a stress test - one that exposes dangerous gaps in how the U. S. protects its critical systems.