Overview
- JPMorgan removed Claude from its approved large-language-model list for employees in Hong Kong on Thursday, restricting staff access to Anthropic’s models in that market.
- The bank linked the step to specific wording in Anthropic’s usage and licensing terms and to broader concerns about regulatory and security exposure for Hong Kong operations.
- Goldman Sachs took a similar step in April, showing a pattern where major banks limit particular AI vendors in sensitive jurisdictions while keeping overall AI programs under strict governance.
- Anthropic suspended access to some models after receiving a U.S. export-control directive on June 13 and is facing a proposed class-action suit in California over subscription usage limits.
- Banks vet AI tools through formal reviews that check data handling, contractual terms, and security controls, a process that is producing uneven, region-specific access rather than blanket bans.