In a recent trip to China, senior fellow Sebastian Mallaby observed that U.S. chip export bans have not stalled Beijing’s AI surge. Chinese firms now rent overseas data‑center capacity to train models, sidestepping U.S. restrictions. Huawei and Hikvision deploy AI for industrial maintenance, while autonomous cars glide through Shenzhen’s streets.

Mallaby argues that the real battle is not raw chip power but AI deployment. China’s open‑source foundation models and tools like OpenClaw demonstrate rapid progress, yet Chinese leaders warn against uncontrolled use. A U.S.‑China agreement on AI safety could mirror the 1968 Nuclear Nonproliferation Treaty, limiting rogue deployments while encouraging collaboration.

The U.S. Senate’s recent bill tightening access to foreign data centers will offer little deterrent; China’s model builders already exploit regional cloud providers. Investors should note that Chinese AI firms are gaining market traction in industrial sectors, potentially eroding U.S. dominance in AI‑enabled manufacturing. A shift from export controls to a global safety framework could reshape competitive dynamics and safeguard against AI misuse.