Club提要：日前，清华大学战略与安全研究中心研究员周波在英国The Ambassador Partnership 网站发表评论文章指出，中美核关系的首要目标是战略稳定。现阶段中国持续推进核力量现代化扩容，但无意与美国追求核均势，始终坚持核力量“小而精”的发展原则，六十余年一贯恪守核武器不首先使用的核心政策。 目前，中美核军控对话存在明显僵局，为管控大国核风险、避免战略误判，中美可在多领域推进机制化合作，包括建立核设施互保规则、暂停破坏性反卫星试验、制度化导弹发射通报，并依托二轨对话规范人工智能军事化核应用，以此夯实双边及全球战略稳定秩序。

打开网易新闻 查看精彩图片

The Ambassador Partnership截图

【文/周波，翻译/北京对话 柳润东】

中美核关系的首要目标是什么？战略稳定。以下六个与核武器相关的议题，将影响未来中美战略稳定。

一、中国不会寻求与美国的核均势，但中国的核武库将进一步增长。北京方面一贯主张其核力量应当“小而精”。美国国防部估计，到2035年中国将拥有约1500枚核弹头。即便这一估计完全准确，中国的核武库仍远不足以与美国逾5000枚核弹头的规模相提并论。

中国扩充核武库的意图何在？回答是：旨在确保自身具备可靠的核二次打击能力，使任何国家在任何情形下都不敢对中国发动核首攻。对于华盛顿而言，这本应被视为一种令人安心的信号：因为华盛顿担忧的恰恰是，如果中国核武器数量过少，在最坏情境下中国可能反而成为率先使用核武器的一方。

二、中国“不首先使用”核武器的政策将继续得到恪守。中国已坚持“不首先使用”原则逾六十年，纵使国力日益强盛，亦未表示过希望放弃这一原则。

北京呼吁所有核武器国家都采纳“不首先使用”政策。这样既能使各国占据道义制高点，又无损其核威慑报复能力。美国以盟国关切为由拒绝作出“不首先使用”承诺。然而，鉴于核打击必将招致大规模核报复，何国敢冒天下之大不韪，对美国盟国实施核突袭？2022年1月，五个核武器国家在联合声明中一致认定：“核战争打不赢，也绝不能打。”既然如此，为何不承诺“不首先使用”？

  1. 三、所谓中国不愿参与双边或多边核谈判其实是一种误解。真正的问题在于各方能够就哪些议题展开对话。近年来，中国不愿就其核力量扩充的规模与速度展开讨论，而美国则拒绝就“不首先使用”问题进行沟通。

僵局可由其他议题率先突破，例如双方可承诺互不以核电站为目标，并互换一份同意不以常规武器打击的核设施清单——印巴之间已采用此种做法。此外，中美双方或许终能就“相互确保脆弱性”（mutually-assured vulnerability）展开讨论；而这一概念，恰恰是美国迄今拒绝承认的。

四、中美两国可约定互不进行反卫星（ASAT）攻击。中、美、俄、印四国均已通过武器试验展示了反卫星能力。两国在拥挤的低地球轨道（LEO）上共运行数以万计的卫星，且数量仍在持续增加。这意味着任何一方都不可能摧毁另一方的全部卫星。然而，即便仅有数颗卫星被击毁，其产生的碎片亦会对己方卫星构成威胁。因此，双方可共同提议一项禁止毁灭性反卫星试验的相互暂停协议。同样，针对高危场景——如卫星近距离接近、碎片碰撞风险、定向能武器在轨试验等——亦应建立及时通报机制。

然而，若美国继续推进“金穹”（Golden Dome）系统——该系统要求在轨道上部署武器平台——则上述合作均无可能。特朗普总统坚称“美国必须、也必将主宰太空”。这不过是一厢情愿的幻觉。任何防御系统都不可能做到100%有效。太空战将是一场全面核战争的前奏或一部分，而核战争中没有赢家。

  1. 五、中美应努力将特定导弹发射的相互通报机制制度化。由于美俄之间、中俄之间已有导弹发射通报的协议，中美若能就相互通报达成一致，将进而促成中、美、俄三方通报机制的形成。2024年9月与11月，中美双方曾就各自的洲际弹道导弹试验相互通报，这是罕见而积极的信任建设范例。

遗憾的是，随后却出现了一个负面例证：2026年8月，美国牵头41国就中国未就一次海射洲际弹道导弹试验提前24小时发布预告一事提出批评（虽未点名中国）。他们声称该试验不符合《海牙行为准则——反对弹道导弹扩散》（HCOC）的要求。这是无稽之谈——中国并非《海牙行为准则》缔约方，并且已就此试验向包括美国在内的相关国家预先通报。

六、既然中美主导人工智能（AI）发展，两国应在人工智能军事应用议题上率先展开对话。 2024年，中美两国领导人曾就核武器指挥控制系统须保持“人在回路”达成共识。然而，迄今为止，尚未就此具体问题进行后续讨论。或许可由非官方的“二轨”专家对话先行破题，推动在以下方面形成共识：禁止完全自主的人工智能核打击算法；将人工智能的作用限定于辅助性威胁评估；并就防范恶意网络攻击对敌方核指挥、控制与通信（NC3）基础设施的破坏，制定共同标准。

中美亦可在以下领域发挥表率作用：战略导弹发射的预先通报；针对人工智能触发的核虚警建立即时热线沟通；就核领域中军事人工智能的应用边界展开交流。联合国安理会五个常任理事国终可就此发表关于人工智能核治理的联合声明，并将“人类对核武器的最终控制”条款纳入《不扩散核武器条约》（NPT）审议进程，从而将双边共识提升为多边法律。

（翻页查看英文）

What can China-US nuclear stability look like?

What is the primary goal of the China-US nuclear relationship? Strategic stability. There are six nuclear-related issues that will matter for that stability in years to come.

  1. 1.China will not seek nuclear parity with the US, although China’s nuclear arsenal will grow further. Beijing maintains that its nuclear arsenal is aimed at being “small and effective”. The Pentagon estimates that China will possess 1500 nuclear warheads by 2035. Even if this turns out to be true, China’s nuclear arsenal will still fall well short of parity with America’s more than 5000 nuclear warheads.

The purpose of that expansion? To ensure that China has a reliable nuclear second-strike capability and that no country dare to contemplate a first strike on China in any circumstances. This ought to be seen as reassuring in a Washington which ought to worry that, if China has only a small number of nuclear weapons, it could be the first to use those weapons in a worst-case scenario.

  1. 2.China’s no-first-use policy on nuclear weapons will continue to be honored. China, having maintained no-first-use for over six decades, has no wish to abandon it now, even if it is becoming ever-stronger.

Beijing has called for all nuclear powers to adopt a no-first-use policy. That would allow them to occupy moral high ground without sacrificing a deterring capability to retaliate. The US has cited the concerns of allies as a reason for not making a no-first-use pledge. But who would venture to launch a nuclear strike on a U.S. ally, knowing it would invite massive nuclear retaliation? In January 2022, the five nuclear weapon states agreed in a joint statement that “a nuclear war cannot be won and must never be fought.” If indeed a nuclear war cannot be won, why not pledge no-first-use?

  1. 3.It is a myth that China is not interested in bilateral or multilateral nuclear talks. The real question is what all parties can agree to talk about. During recent years, China has been reluctant to talk about the size and speed of its nuclear buildup, while the US has refused to talk about no-first-use. That ice can be broken by exploring other matters such as non-targeting of nuclear power plants, and they can exchange a list of nuclear facilities that they agree not to target with conventional weapons, as India and Pakistan are doing. Also, perhaps they can finally talk about mutually-assured vulnerability, something that the US has so far refused to recognize.

  2. 4.China and the US could agree not to conduct anti-satellite attacks (ASAT). Both of them, along with Russia and India, have demonstrated anti-satellite capabilities through weapon tests. China and the US combined have tens of thousands of satellites sharing congested low earth orbits (LEO) and more are coming. This means neither side can destroy all the satellites of the other side. But the debris from even a few destroyed satellites could pose risks to their own satellites. Therefore, they could jointly propose a mutual moratorium on destructive tests. Similarly, timely notification procedures should be set up for high-risk scenarios, such as satellite close approaches, debris collision risks, and in-orbit testing of directed energy devices.

Yet none of this is possible if the US continues with the Golden Dome system which requires placing weapon systems in orbit. President Trump believes “we must have American dominance in space. And so we will”. This is a hallucination. No defense system is 100% effective. A war in space would be a prelude to or part of an all-out nuclear war in which there cannot be any winners.

  1. 5.China and the US should work towards institutionalizing mutual notification of specified missile launches. Because there is already agreement between Russia and the US on mutual notification and between China and Russia on missile launch notifications, an agreement on mutual notification between China and the US can lead to a trilateral notification agreement. In September and November, 2024, Beijing and Washington notified each other of intercontinental ballistic missile (ICBM) tests. This was a rare yet good example of confidence-building. Alas, it was followed by a bad example: in August 2026, the US led a group of 41 countries in criticizing China (without naming it) for not providing 24 hours’ advance notice of a sea-launched ICBM test into the South Pacific. They complained that the test was not in line with the Hague Code of Conduct Against Ballistic Missile Proliferation (HCOC). This was nonsensical because China is not a party to the HCOC and China did give advance notice to relevant countries, including the US.

  2. 6.Since China and the US dominate AI development, they should take the lead in talking about the application of AI in the military field. In 2024,the leaders of China and the United States agreed to have human oversight over nuclear command & control systems. But to date there have been no follow-up discussions on this specific issue. Perhaps a non-governmental experts’ dialogue could help to achieve agreement on the following: a ban on fully autonomous AI nuclear strike algorithms, restricting AI to advisory threat assessments, and setting shared standards against malicious cyber disruption of rival nuclear command, control and communications (NC3) infrastructures.

China and the US can also take the lead on pre-notification of strategic missile launches, immediate hotline communication on AI-triggered nuclear false alarms, and exchanges on military AI boundaries in nuclear domains. Eventually, the five permanent members of the UN Security Council could make a joint statement on AI nuclear governance and could integrate human nuclear control clauses into the Treaty on the Non-Proliferation of Nuclear Weapons review process, elevating bilateral agreement into multilateral law.

打开网易新闻 查看精彩图片

本文系观察者网独家稿件，文章内容纯属作者个人观点，不代表平台观点，未经授权，不得转载，否则将追究法律责任。关注观察者网微信guanchacn，每日阅读趣味文章。