JPMorgan Chase has restricted its employees in Hong Kong from accessing Anthropic’s AI language models, including Claude, echoing a similar move by Goldman Sachs earlier this year. According to multiple sources familiar with the situation, the decision stems from the specific terms of Anthropic’s licensing agreement with the bank, which limit usage in regions including Greater China—the designation encompassing Hong Kong.

This restriction prevents Hong Kong-based staff from selecting Claude models via the firm’s internal approved list of large language models (LLMs). The move comes amid growing scrutiny over the deployment of advanced AI technologies outside the United States, particularly in regions where regulatory and security concerns are heightened.

Anthropic’s models, along with Western AI platforms such as OpenAI’s ChatGPT, are officially banned in mainland China under the country’s internet censorship framework, often referred to as the “Great Firewall.” Unlike mainland China, Hong Kong has generally operated without the same level of direct Chinese technological censorship, although companies providing AI services have imposed their own limitations on usage within the territory.

International organizations have managed to bypass geographic restrictions through global contracts and by hosting AI-related activities outside of mainland China, allowing some access despite official bans. However, US-based AI developers remain cautious about usage within China and Greater China due to concerns over “distillation,” a process by which local entities could exploit extensive use of foreign AI models to develop local versions.

The decision to block access in Hong Kong is seen by some industry observers as potentially detrimental to the city’s efforts to reestablish itself as a global financial center, where rapid adoption of AI tools—particularly for tasks like coding—has played an increasingly important role.

The restrictions follow broader national security concerns voiced by US officials, who recently requested Anthropic to limit access to its advanced AI model, known as Fable or Able, for foreign users. Government authorities and financial executives have expressed apprehensions about Anthropic’s Mythos model, citing its potential to exploit cybersecurity vulnerabilities within the international financial system.

JPMorgan Chase has declined to comment on the matter, and Anthropic has not responded to requests for comment. The ongoing restrictions reflect the complex intersection of technological innovation, international regulatory environments, and geopolitical sensitivities surrounding the use of artificial intelligence in global financial markets.