随著中国领导人即将访问华府，人工智慧议题成为美中对话的重中之重。当前美国社会对失控 AI 的生存风险与经济冲击日益担忧，业界高层亦呼吁放慢前沿创新步伐。然而，若仅由美国实验室单方面暂停研发，将无法真正消除风险，唯有配合中国的自我克制才具有实质意义，否则可能重创美国的战略优势。

确保中国同步放缓 AI 发展面临两难途径：第一条路是透过外交途径协商共同减速，但北京往往将此视为冷战遏制手段，且存在极高的验证与违约风险；第二条路则是全面升级出口管制，防堵云端与海外子公司漏洞，切断先进晶片与半导体设备供应，以强制手段剥夺中国训练前沿模型所需的运算资源。

尽管更严格的技术管制能强化美方谈判筹码并遏制对手进展，但也极可能引发中国在稀土等领域的反制，导致双方脆弱的缓和关系破裂，陷入更剧烈且不受约束的冷战对抗。尽管目前政治层面仍有争议与阻力，但美国如何平衡 AI 安全与战略竞争，并有效应对中国挑战，已是无法回避的核心课题。

As Chinese leadership prepares to visit Washington, artificial intelligence has emerged as a top agenda item amid mounting US anxieties over economic disruptions and existential risks posed by rogue AI models. While prominent AI leaders advocate for pacing frontier innovation, any unilateral pause by American labs would be strategically self-defeating without reciprocal restraint from China, which is trailing closely behind in technological capabilities.

The United States confronts two difficult pathways to secure Chinese compliance: either negotiating a mutual slowdown through diplomatic channels, or mounting an aggressive export-control offensive to choke off China's access to advanced chips, semiconductor equipment, and compute infrastructure. While diplomacy faces severe trust and verification deficits, coercive export controls aim to enforce a slowdown unilaterally to preserve America's technological lead.

However, relying on intense technological coercion risks severe blowback, including Chinese retaliation through rare-earth export curbs and the potential collapse of fragile bilateral detente into an unconstrained cold war. Even as political leaders debate the necessity of pausing AI, the dilemma of balancing safety concerns with maintaining a decisive strategic edge over Beijing remains an urgent, unavoidable challenge.