JPMorgan Chase has suspended access to Anthropic’s AI models for its Hong Kong employees, intensifying scrutiny over the use of advanced artificial intelligence outside the United States. Staff at the bank’s operations in the Asian financial hub can no longer select Claude, Anthropic’s large language model, from the internal list of approved AI tools, according to multiple sources familiar with the decision.

The restriction stems from licensing terms in Anthropic’s agreement with JPMorgan that appear to prohibit use of its technology in Greater China, encompassing Hong Kong. This mirrors an earlier move by Goldman Sachs, which also barred its Hong Kong-based bankers from utilizing Anthropic’s AI after interpreting the company’s usage policies as excluding the region.

The limitation on access to cutting-edge AI technology presents a challenge to Hong Kong’s ambitions of reestablishing itself as a leading international financial center. Other global markets have rapidly adopted such tools, particularly for coding and productivity enhancements. Unlike mainland China, where Western AI models such as OpenAI’s ChatGPT and Anthropic’s Claude are blocked under the country’s extensive internet censorship system known as the Great Firewall, Hong Kong has traditionally remained free from these restrictions. However, the usage limitations imposed by AI providers themselves now curtail access in the city.

While direct connections from Hong Kong to these models are restricted, some multinational firms have previously circumvented regional barriers by entering global contracts and routing AI activity through servers outside China. Nevertheless, US-based AI developers express concern about their technology being used in China due to fears of “distillation,” a process by which proprietary models could be reverse-engineered or used to train local versions.

National security has also become a focus in recent weeks. Anthropic was compelled to suspend access to its advanced Fable AI model after US government officials requested tighter controls on foreign access, citing risks related to cybersecurity threats in the global financial sector. The AI lab’s Mythos model had drawn particular scrutiny for potential vulnerabilities.

JPMorgan declined to comment on the matter, while Anthropic did not respond to requests for comment. The developments reflect growing tensions around the deployment of powerful AI tools amid geopolitical and regulatory concerns in key financial centers.