Dr Ruby Tong is a university innovator and technology strategist bridging academic research, deep tech commercialisation, and AI governance.
The market realignment over the past week starkly captures the self-defeating logic of America’s technology containment strategy.
The momentum began on June 15, when the shares of Chinese artificial intelligence pioneer Zhipu AI, which trade in Hong Kong under Knowledge Atlas Technology, surged by 48 per cent intraday after it announced the open-source release of its GLM-5.2 foundation model. The rally continued this week, with Zhipu’s shares trading at an intraday high of HK$2,980 (US$380) in the morning of June 22, briefly driving market capitalisation past HK$1.2 trillion – a 25-fold appreciation on its January debut.
This massive capital migration followed a sudden regulatory intervention on June 12, when the US Commerce Department ordered California-based Anthropic to block foreign access to its new Fable 5 and Mythos 5 models. Because Anthropic could not practically verify user nationalities in real time, it had to shut down both flagship models globally for everyone, lock out its own non-American staff and downgrade affected subscribers to an older model.
Although Washington blamed a “jailbreak” security vulnerability, Anthropic’s technical review showed the exploit involved only minor flaws that already exist in other publicly available models.
The situation is profoundly ironic. Anthropic supported Washington’s export controls, actively lobbying for stricter cross-border limits. Having helped build this regulatory framework, the company is now trapped by it, exposing a fundamental flaw in techno-nationalist logic. Washington treats software models like physical nuclear components, yet software cannot be embargoed like hardware. Because code moves instantly, denying access simply drives other nations to develop alternative technologies more quickly.
This architecture did not emerge overnight. Since October 2022, Washington has systematically restricted Nvidia’s advanced silicon, expanded its blacklist and globalised these controls by bringing foundational open-source models under its restrictive umbrella.
