How worried should we be about rogue AI agents?

As frontier models commit their ‘first felony’, fears mount about AI’s ‘unchecked proliferation’