JPMorgan blocks Claude access for Hong Kong staff

JPMorgan has blocked access to Anthropic’s chatbot Claude for employees in its Hong Kong office. People familiar with the matter said the restriction was implemented recently and prevents use of Claude on company devices or accounts.

The bank is reviewing potential risks linked to third-party generative AI tools, including how external models handle client and proprietary data and whether they meet local regulatory expectations for data residency and auditability.

JPMorgan did not provide a detailed public explanation for the restriction. Financial firms commonly limit employee use of externally hosted chatbots unless vendors meet security and contractual requirements. Some institutions run their own internal AI systems or sign contracts that specify data storage, retention and confidentiality terms before permitting business use.

Anthropic offers Claude under different commercial arrangements and security settings for enterprise customers. Use of public interfaces by individual employees can create uncertainty about where inputs are stored and whether those inputs are used to improve models.

Hong Kong regulators have issued guidance on vendor management, controls over customer data and transparency around model training and outputs. It is not clear whether the Claude restriction is temporary while JPMorgan completes its review or a longer-term policy applying to the Hong Kong office.

Employees can face different access to external AI tools depending on location and whether a use case has been approved by the bank’s technology, risk and compliance teams. Firms typically require vendor risk assessments, contractual guarantees about data handling and technical safeguards such as encryption and logging before allowing external models to process client or confidential information.