JPMorgan Chase (NYSE:JPM) has become the latest Wall Street bank to cut off employee access to artificial intelligence tools from Anthropic in its Hong Kong operations. According to a report in the Financial Times, the firm removed Anthropic’s Claude large language models from an internal list of approved AI systems available to staff in the city.
The decision stems from the specific language in Anthropic’s licensing agreement with the bank, which does not appear to authorize use in Hong Kong.
Bankers there can no longer select the models from a drop-down menu of permitted tools within JPMorgan’s internal systems.
This move reflects heightened caution among financial institutions regarding the deployment of advanced AI outside the United States, particularly in regions tied to complex geopolitical dynamics.
The action follows a nearly identical step taken by rival Goldman Sachs earlier this year.
In April, Goldman Sachs also withdrew access to Anthropic’s Claude models for its Hong Kong-based employees after lawyers conducted a strict review of the licensing contract.
That earlier restriction similarly limited the AI tools available to investment bankers in the Asian financial center.
Both decisions highlight a pattern of US financial firms tightening controls on AI usage in Hong Kong amid broader concerns over technology transfer, data security, and regulatory compliance.
Western-developed AI systems face significant barriers in mainland China, where access is generally restricted.
Hong Kong has operated under a somewhat more flexible regime, yet U.S. companies have increasingly imposed their own limitations to align with national security considerations and contractual obligations.
The restrictions come at a time when US authorities have intensified scrutiny of advanced AI exports and usage.
Government actions, including orders affecting certain models from providers like Anthropic, underscore worries about potential misuse by entities in countries of concern.
Financial institutions, which handle highly sensitive client data and operate under strict regulatory frameworks, appear particularly attuned to these risks.
Removing access prevents any possibility of inadvertent violations related to licensing terms or cross-border data flows.
For JPMorgan and Goldman Sachs employees in Hong Kong, the change means reduced options for AI-assisted tasks such as document summarization, code assistance, or research support that Claude models previously provided.
While banks continue to invest heavily in artificial intelligence to boost efficiency and analytical capabilities, they are also prioritizing compliance and risk mitigation.
Internal approved lists of tools serve as a key control mechanism to ensure employees only use vetted systems that meet legal and security standards.
Industry observers note that such moves could influence other global banks operating in Asia.
As AI capabilities advance rapidly, financial firms must navigate a patchwork of international rules, corporate contracts, and geopolitical tensions.
The focus on licensing language in particular suggests that companies like Anthropic are embedding geographic restrictions into their enterprise agreements to manage regulatory exposure.
These developments illustrate the practical challenges of rolling out powerful AI technologies on a global scale.
Banks must balance the productivity gains offered by tools like Claude against the need to safeguard against compliance breaches, data leakage risks, and potential national security implications.
JPMorgan Chase’s decision reinforces a growing trend of selective AI adoption, where access is granted or withheld based on location-specific legal and contractual factors rather than uniform global rollout. As more institutions review their AI partnerships, further adjustments in regions like Hong Kong appear likely.