JPMorgan Chase has instructed staff in its Hong Kong office that Anthropic’s Claude models are no longer available for selection from the bank’s internal list of approved large language models, citing three people familiar with the matter.
The restriction is tied to specific wording in Anthropic’s usage terms as set out in the licensing agreement between the bank and the AI developer. According to those familiar with the decision, the language in the contract prompted the bank to remove Claude from the approved-model choices for employees based in Hong Kong.
Bank officials applied the change to the internal list of authorized models used by staff in the region, meaning employees in Hong Kong will not be able to route work to Claude through the bank’s established channels for third-party AI tools. The action follows a comparable move by another large Wall Street institution earlier in June.
Separately, access to certain Western AI services is restricted in mainland China under the country’s internet controls, which prohibit platforms such as OpenAI’s ChatGPT and Anthropic’s Claude from operating inside the mainland. Hong Kong has generally been treated differently from mainland China with respect to those restrictions, but this corporate policy applies specifically to JPMorgan’s internal approvals and licensing considerations for its Hong Kong staff.
The development comes amid wider industry adjustments to AI access: last week Anthropic temporarily suspended access to its advanced model Fable after receiving a request from the U.S. government to limit access for foreign citizens on national security grounds. That suspension and the licensing-language concerns cited by the bank are separate but contemporaneous developments affecting how firms decide which models to permit for employee use.
The bank’s decision reflects a contractual and compliance-driven step rather than a broader public policy change in Hong Kong. It affects the subset of employees who rely on the institution’s approved list of third-party large language models for work-related tasks and underscores how licensing terms and government requests can influence corporate AI usage rules.
Summary of developments
- JPMorgan has removed Anthropic’s Claude models from the approved-model list for employees in its Hong Kong office, based on language in the licensing agreement.
- The move parallels a similar action taken by another major bank earlier in June.
- Separately, Anthropic suspended access to its model Fable after a U.S. government request to restrict foreign citizen access for national security reasons.