Goldman Sachs employees in Hong Kong were cut off from using Anthropic's Claude model a few weeks ago — and the restrictions go further than just the office. The geo-block applies even to staff who fly into Hong Kong from overseas. They can't access Claude while on the ground. Affected most are software engineers using the model to write code and build financial models.
Goldman Sachs interpreted its contract with Anthropic strictly and, after discussions, decided no Hong Kong employee should use any Anthropic product. An Anthropic spokesperson told the Financial Times that Claude has never been officially "supported" in Hong Kong, but declined to elaborate. The restriction doesn't extend to other AI providers — employees still have access to ChatGPT and Gemini through the bank's internal platform.
Why the clampdown? The core concern for US AI companies is "distillation" — local organizations training their own models by aggressively calling foreign model outputs. ChatGPT and Claude are already blocked in mainland China, but Hong Kong was previously a gray area, with usage boundaries set by each US company. Hong Kong is the investment banking and cross-border transaction hub for most global banks in Greater China; if staff there can't use the most capable models, they risk falling behind. Meanwhile, Chinese AI rivals continue to advance — DeepSeek V4 now offers 1 million tokens of context as standard. The Financial Times reports it's unclear whether other banks or firms have implemented similar restrictions.