On 17 September 2027, a Pugwash delegation was privileged to be invited to participate in the Linhu Think Tank Distinguished Lecture Series on International Arms Control and Global Security, held under the auspices of the School of International Studies at Peking University.
The meeting produced a rich and fruitful exchange of ideas, centred first on nuclear risk reduction as a more feasible near-term alternative to formal arms control in the US–China context, and second on how artificial intelligence and transformational technologies are disrupting the challenges of strategic stability, international security, and arms control.
- The starting point for discussions was that deliberate war between nuclear rivals is irrational and undesirable and thus the highest risk of nuclear weapons use lies in miscalculations, misunderstandings or mistakes provoking a cycle of inadvertent escalation.
- Participants presented China’s No-First-Use (NFU) policy as deeply embedded in Chinese strategic thinking, rooted in a tradition of ‘strike only after the enemy has struck’. Chinese analysts believe that the credibility of China’s No-First-Use pledge is reinforced by the high reputational and audience costs that any violation would carry with China’s allies and across the Global South.
- There was some pushback over how the NFU commitment fits with China’s significant nuclear modernisation and that deeper explanation of this disjuncture should be a priority for Chinese policymakers.
- All the same, concepts and initiatives that US policy experts have raised can complement exploring what a No-First-Use world may look like: ‘no early second use’ and ‘zero ballistic missiles’ imply greater de-alerting of nuclear forces, thereby reducing unwanted surprises and providing stability. Implications of re-deployment of intermediate range nuclear missiles were also discussed, as they may increase the risk of nuclear weapons use in the region.
- A significant gap remains between US and Chinese strategic cultures: the United States tends to draw lessons from past crises and periods of tension, while China’s outlook is shaped more by its understanding of core interests and the balance of power. Questions were also raised about China’s evolving nuclear strategic thinking and whether it is converging with US deterrence and readiness posture. Participants agreed that exploring this issue has important implications for crisis management and for understanding escalation dynamics.
- On disruptive technologies, there was deep concern over the absence of guardrails, underscored by the seriousness of cybersecurity events in the US over recent months. Noting the different levels of AI development between the US and China, participants agreed that without an industry-led slowdown or governmental pact, capabilities may outrun governance options, verification possibilities, and, ultimately, human control.
- One area of research could be to cooperatively map how states may integrate AI to match specific nuclear postures, with the goal to identify risk-minimizing deployments of AI in nuclear systems.
- Participants believed there was potential for US–China cooperation on AI safety, specifically in the areas of incident reporting, biosecurity risks, and applying arms-control thinking to AI governance.
- Overall, participants highlighted the absence of strong communications and dialogue in the US-China relationship and among the N-5 in general as a key challenge. Several participants felt that convening a new global nuclear summit would demonstrate China’s responsibility as a global power.
- Participants agreed that some guardrails to stabilize US-China relations were desirable, and risk reduction tools for both the nuclear and AI domains could be pursued under four general categories:
- identifying transparency measures that do not compromise nuclear postures or deterrence relations
- creating direct communication channels at different political and military levels, including incident reporting for AI events
- pursuing deeper strategic dialogue to promote mutual understanding on the shared threats and risks
- stabilizing measures to minimize surprises
