Bankers at JPMorgan Chase and Goldman Sachs in Hong Kong were abruptlyabruptly/əˈbrʌptli/L3突然地；出乎意料地in a sudden and unexpected manner, often with little or no warning barred from using Anthropic's frontier AI models in April and May, following strict US export controls. This enforced disconnection threatens to undermine Hong Kong's competitiveness as a financial hub, given global rivals' rapid integration of such models. The Financial Times has warned that barring access to the world's most sophisticated AI constitutes a significant threat to the territory's revival as an international financial centre. Consequently, the decision by these American banks, however compliant with US regulations, may inadvertentlyinadvertently/ˌɪnədˈvɜːrtəntli/L3无意地；非故意地without intention or awareness; accidentally accelerate the erosionerosion/ɪˈroʊʒən/L3侵蚀；逐渐削弱the gradual destruction or weakening of something, such as a competitive position or value of Hong Kong's comparative advantage.
A US Commerce Department directive this month gave Anthropic just 90 minutes to shut off access to its most advanced models, citing national security concerns. This export control measure effectively cuts off global access, including for Anthropic's own foreign employees. If foreigners cannot use America's frontier models, most will switch to Chinese alternatives, which are far cheaper. DeepSeek's V4 Pro, for instance, is reportedly 60 times cheaper than Anthropic's Fable 5 and topped a global efficiency ranking.
Chinese AI models are not merely 'good enough'; their performance is increasingly on par with US counterparts, as noted by geoeconomics fellow Agathe Demarais. Moreover, they are open source, allowing clients to customise them without fear of sudden regulatory shutdowns. Demand for Chinese models has already overtaken US models on OpenRouter, according to Fortune's Nicholas Gordon. Indeed, the US ban may end up vindicatingvindicating/ˈvɪndɪkeɪtɪŋ/L3证明…正确；为…辩护proving that someone or something is right, justified, or free from blame China's broader move towards tech self-sufficiency.
In contrast, Europe is alarmed; former French minister Bruno Retailleau warned that reliance on foreign technology leaves a nation vulnerable to being 'unplugged overnight'. British MP Alistair Carns noted that the restrictions immediately affected British researchers and hospitals piloting the models. Meanwhile, Anthropic's CEO Dario Amodei, who had warned the US government about his models' dangers, may have been too convincing. As Patrick Boyle wrylywryly/ˈraɪli/L3挖苦地；讽刺地in a way that expresses dry, ironic, or mocking humor observed, telling authorities you have built a weapon and then being shocked when they treat it as one is bad corporate management.
