JPMorgan Chase has removed Anthropic’s Claude from its list of approved large language models for employees in Hong Kong, following a similar move by Goldman Sachs in April. The restrictions come as Wall Street banks reassess the risks of using advanced AI tools in sensitive markets, particularly where licensing terms and geopolitical tensions intersect.
The decisions were reportedly driven by contract language in Anthropic’s licensing agreement that excludes usage in Greater China, including Hong Kong. While Hong Kong-based Goldman staff can still access other AI services such as OpenAI products and Google’s Gemini through internal platforms, the ban remains specific to Claude.
These moves highlight a growing trend: AI adoption in finance is becoming regionally fragmented. Banks that aggressively invest in generative AI for automation, compliance, and research are now forced to balance productivity gains with strict legal and data-sovereignty requirements. JPMorgan, which already uses AI across internal functions, is adjusting tool access where policy risks appear higher—even as it continues broader AI integration elsewhere.
The backdrop includes rising U.S.-China tensions over technology transfer and fears that foreign users could accelerate domestic AI development through model distillation. Anthropic itself recently suspended certain model versions after a U.S. government export-control directive, and faces a class-action lawsuit over paid plan limits. These developments reinforce the strategic risk now tied to AI deployment, with analysts questioning whether this fragmentation will extend to other frontier AI systems and reshape global financial centers.
