Anthropic’s powerful AI model, Mythos 5, gets government go-ahead for “trusted” US organizations after a two-week ban.