在9月24日华盛顿峰会上，唐纳德·特朗普与习近平达成共识，宣布建立旨在通报人工智能突发事件的双边“沟通渠道”，并由美国财长斯科特·贝森特与中国副总理何立峰牵头于11月前举行第二轮AI对话。美方官员将其类比为冷战时期美苏防范核冲突的热线机制，寄望以此应对超级模型失控风险。然而，资深外交界人士对该机制的实用性普遍持怀疑态度：回顾1998年美中因台海危机设立元首热线后，1999年北约误炸中国驻南联盟使馆时江泽民整整一周拒接克林顿总统电话；2001年中美南海撞机事件后美方在12小时内无法联络到中方官员；2008年设立的军方加密专线不仅需提前48小时预约，且在2009年南海对峙以及2023年间谍气球事件中均遭中方直接拒接，呈现出危机爆发时中方决策系统往往陷入静默的常态。

中美热线频遭搁置的根本原因在于两国战略文化与权力感知的深刻错位。不同于冷战时期美苏视彼此为军事均势并拥有明确防范核毁灭的共同利益，中国自视处于相对弱势地位，因而将美方主导的安全护栏视作限制其战略行动自由的规锁手段。此外，面对突发事件，中方官僚体系倾向于保持通讯静默，以便高层在委员会闭门会议中统一口径并研判情报，同时通过故意拖延时间来最大限度抵消美方追求迅速解决危机的谈判优势。尽管热线在领导人国事访问、2008年全球金融危机协调以及2020年大选期间有效澄清了美方军事意图，但在不可预测的高烈度危机初期，传统的即时语音通话几乎完全失效。

为了避免AI热线沦为形式主义摆设，专家建议必须吸取历史教训并重构沟通架构：首先，应放弃同步语音通话，转为依托传真或加密文本通信，以便为中方提供内部审议时间并规避语调误读与翻译偏差，同时适应军队高层频繁的人事洗牌；其次，卡耐基国际和平研究院学者建议严格界定触发热线启动的具体风险阈值，并参照1998年《中美海上军事安全磋商协定》（MMCA）每年两次的例会机制，由两国的AI领域专业技术官僚设立独立工作组，专司日常算法漏洞与无国家背景恶意攻击等技术性风险的信息互换，从而在两国领导人11月重逢前将安全护栏落到实处。

At the September 24th Washington summit, Donald Trump and Xi Jinping formally agreed to institute a bilateral emergency communication mechanism for critical artificial intelligence incidents, tasking US Treasury Secretary Scott Bessent and Chinese Vice-Premier He Lifeng with convening a second dialogue round by November. While Washington drew analogies to Cold War crisis hotlines that averted nuclear conflict following the 1962 Cuban missile crisis, seasoned diplomats remain deeply skeptical. Precedent underscores persistent functional paralysis: established after the 1996 Taiwan missile crisis, the 1998 presidential hotline went unanswered for a week by Jiang Zemin following the 1999 Belgrade embassy bombing; American officials were stonewalled for 12 hours after the 2001 Hainan spy plane collision; and the 2008 Pentagon-Zhongnanhai military line—cumbersomely requiring 48 hours advance notice—was deliberately ignored during maritime confrontations in 2009 and the 2023 spy balloon incident.

This repeated breakdown stems from asymmetric power perceptions and institutional crisis doctrines. Unlike the Soviet Union, which negotiated as an acknowledged military peer, Beijing perceives itself as the structurally disadvantaged power, viewing American-sponsored operational guardrails as strategic containment ploys designed to constrain Chinese flexibility. Operationally, China’s centralized political apparatus systematically "goes dark" during emerging crises to preserve strategic ambiguity, gather internal intelligence, and forge consensus through opaque committee deliberations, deliberately exploiting procedural delays to dilute Washington's urgency. Although hotlines facilitated administrative protocols during the 2007–09 global financial crisis and de-escalated military tensions ahead of the 2020 American elections, real-time voice diplomacy consistently fails at the inception of acute emergencies.

To construct a functional AI crisis architecture, experts argue the protocol must diverge structurally from previous telephone links. Security analysts advocate replacing voice channels with encrypted written facsimile transmissions to eliminate auditory tone misinterpretations, facilitate technical translation, and provide administrative buffer time accommodating elite personnel turnover. Furthermore, researchers at the Carnegie Endowment recommend negotiating precise objective activation thresholds alongside establishing technical working groups modelled after the 1998 Military Maritime Consultative Agreement (MMCA). Such dedicated technical tracks, staffed by specialized computer scientists, would manage routine bilateral exchanges on rogue models and non-state cyber actors, providing durable institutional guardrails before heads of state reconvene in November.

Source: In an AI crisis, would China answer the phone?

Subtitle: The Americans want a hotline but the Chinese have been cool in the past

Dateline: Oct 1st 2026\n