JPMorgan Chase bans Hong Kong employees from accessing Anthropic, tightening compliance measures after Goldman Sachs

According to Dongcha Beating Monitoring, citing informed sources reported by the UK’s Financial Times, JPMorgan has stopped its Hong Kong branch employees from accessing AI models developed by Anthropic. The report says that employees at the Hong Kong branch are currently unable to select the Claude model from the in-house approved large language model dropdown list. This is yet another Wall Street institution in Hong Kong restricting the use of Claude models, following similar actions taken by Goldman Sachs earlier this year.

Insiders said that JPMorgan’s decision to disable the Claude model is mainly based on the restrictive descriptions in Anthropic’s usage terms in the licensing agreement. Previously, Goldman Sachs also made a strict interpretation of Anthropic’s service terms. The service terms explicitly exclude Greater China (including Hong Kong) from the scope of the service.

American AI companies have implemented usage restrictions for Greater China, partly to guard against the “distillation” risk—i.e., Chinese local teams training their own autonomous large models by making high-frequency calls to cutting-edge models. At present, Western advanced models such as ChatGPT and Claude cannot be directly accessed in Hong Kong. Multinational companies typically bypass geographic restrictions by signing global contracts and hosting computing power abroad. However, the restrictions imposed by Wall Street banks show that the contractual barriers caused by geopolitical conflicts are filtering through to multinational financial institutions. Since the global financial industry has widely adopted cutting-edge AI models for core businesses such as coding, restricting employees’ access may negatively affect Hong Kong’s competitiveness as an international financial center.