Goldman Sachs is banning access to Anthropic’s Claude for bankers in Hong Kong, according to a source cited by the Financial Times.
In the Tuesday report, the FT claimed that one person familiar with the move said Goldman acted after adopting a strict interpretation of its agreement with Anthropic following discussions with the Silicon Valley company, a decision that has left employees in Hong Kong unable to access Claude models for weeks, either directly or through the firm’s internal AI platforms.
According to the FT, Goldman’s move is specific to Anthropic and does not affect contracts with other AI vendors such as OpenAI.
While AI tools like ChatGPT and Claude are banned in mainland China through what it calls the Great Firewall, Hong Kong has remained more open, with usage restrictions generally determined by the AI companies rather than Chinese regulators.
This is because American AI companies are increasingly cautious about activity in China, in part because of concerns over “distillation,” where local developers could use foreign models at scale to train new competing tools, which can be seen in recent public accusations.
Although no evidence has been publicly presented to support the claim, OpenAI last year accused Chinese competitor DeepSeek of using its models to train another system. The US government this month also alleged China was engaged in “industrial-scale” theft of US AI companies’ intellectual property, an accusation the Chinese embassy in Washington dismissed as “pure slander.”
Even though Hong Kong continues to serve as the primary investment banking and finance hub for Greater China for most global banks, which use the territory to coordinate cross-border activity spanning trading, mergers and acquisitions, and share sales, Goldman’s move to limit staff access to Claude may pose a setback to its ambition to rebuild as a financial and knowledge hub, especially if employees using the model for coding and financial work fall behind counterparts with broader AI access.
David Adubiina
