“Reid Hoffman suggests a regulatory framework where AI companies must generate reports on key concerns, with auditors required to alert the government if metrics worsen quickly.”