CISA’s ChatGPT incident exposes a growing AI governance gap as enterprises struggle to define who is accountable for actions taken by AI agents.