人工智能风险相关术语研究（一）

1第1期（总第38期） 人工智能风险相关术语研究（一）2025 年第 1 期（总第 38 期）REPORTGlossary Research on Artifi cial Intelligence Risks (Part I)01第1期（总第38期）人工智能风险相关术语研究（一）Glossary Research on Artifi cial Intelligence Risks (Part I)CISS人工智能与国际安全项目术语工作组By CISS Working Group on Artifi cial Intelligence Glossary编者按：2025年2月13-14日，清华大学战略与安全研究中心与美国布鲁金斯学会在慕尼黑举行第12轮“中美人工智能与国际安全二轨对话”。在术语讨论环节，双方共同提出20个人工智能风险相关的概念术语进行探讨交流，并就其中9个术语给出了各自的具体解释，以此增进彼此对人工智能潜在安全风险的认知和理解。On February 13-14, 2025, the Center for International Security and Strategy (CISS) of Tsinghua University and the Brookings Institution held the 12th round of the“China-U.S. Track 2 Dialogue on Artifi cial Intelligence and International Security” in Munich. During the Glossary Session, both sides proposed 20 AI risk-related terms for exchange, and then provided their respective definitions for nine of these terms to enhance mutual understanding of the potential security risks posed by artifi cial intelligence.●灾难性风险灾难a性风险b特指破坏力极强、远超常规应对能力的极端事件，其核心特征可归纳为三点：破坏烈度极高、影响范围广泛、持续时间漫长。这类风险既可能源自自然力量（如超级火山爆发），也可能源于人为因素（如核战争），其破坏性不仅会摧毁社会基础架构，更将a 灾祸造成的苦难;灾祸。《汉语大词典》（2007版），上海：上海辞书出版社；天灾人祸造成的严重损害和苦难。《辞海》（第6版），上海：上海辞书出版社，第5582页。b 可能发生的危险。《汉语大词典》（2007版），上海：上海辞书出版社。02国际战略与安全研究报告引发跨地域、跨代际的深远影响。在人工智能领域，该类风险可能体现为两种形态：其一，系统因复杂架构产生不可预测的突发行为；其二，技术遭恶意滥用导致大规模危害。其形成机制通常源于两大矛盾，即技术迭代速度超越人类监管能力，或现有治理体系无法有效约束技术发展使其符合社会整体利益。●Catastrophic riskCatastrophic risk refers to events with exceptional destructive potential that exceeds ordinary response capacities, characterized by significant magnitude, geographic scope, and temporal persistence. Such risks, whether natural or anthropogenic in origin, disrupt fundamental societal structures and produce multi-generational consequences that transcend jurisdictional boundaries.In the artificial intelligence domain, catastrophic risks emerge when systems potentially inflict widespread harm through either complexity-induced emergent behaviors or deliberate misuse. These scenarios typically develop through two primary mechanisms: technical systems evolving beyond effective human oversight, or governance frameworks proving insufficient to ensure responsible development aligned with broader societal interests.●生存风险作为风险谱系中最严峻的类别，生存c风险直接威胁人类物种存续，可能彻底阻断文明发展轨迹。相较于灾难性风险，这类风险（如小行星撞击、超级病原体或失控AI）具有不可逆特性，能够彻底消弭文明复苏的可能，根本上颠覆人类赖以生存的基础条件。 在人工智能领域，该类风险可能出现的情景包括：系统具备超越c 活着，活下去；存在；生活。《汉语大词典》（2007版），上海：上海辞书出版社；活着；保存生命。《辞海》（第6版），上海：上海辞书出版社，第3960页。03第1期（总第38期）人类认知边界的智能水平、文明演进方向出现根本性偏移、人类在关键领域丧失控制权、自主系统脱离监管破坏战略稳定，或有效治理框架之外运行的自主能力所导致的战略不稳定。●Existential riskExistential risk refers to threats capable of causing human extinction or permanently curtailing humanity’s developmental trajectory, representing the terminal category within risk taxonomies. Such threats, whether cosmic, ecological, biological, or technological, are distinguished from catastrophic risks by their irreversible elimination of recovery pathways, potential to extinguish humanity entirely, and capacity to fundamentally alter parameters necessary for continued human existence.In a