Rand兰德：2025年人类能否设计出可靠的实用安全措施来对抗人工超智能代理？（英文版）

Dramatic advances in frontier artificial intelligence (AI) raise the question of whether humans can design practical safeguards to protect U.S. critical and digital infrastructure against attacks from a future artificial superintelligence (ASI). Some argue that a superintelligent agent would have such mastery and power over physical space and cyberspace that it could thwart any effort by humans to prevent it from undermining the confidentiality, integrity, and availability of our information.