“Framing emphasizes risk mitigation and the potential threat of uncontrolled AI agents, invoking mild caution without overt alarm.”