Advancing AI Safety: A Structured Path to Impact | Manifund