“Framing emphasizes uncertainty and failure of oversight, creating a tone of worry regarding AI safety risks.”