JPMorgan reportedly removes Claude from approved tools for Hong Kong staff

JPMorgan Chase has reportedly limited the use of Claude, an artificial intelligence assistant developed by Anthropic, for employees in Hong Kong. The change, described in a Yahoo Finance report published Tuesday, indicates the bank is tightening which AI tools are considered “approved” for certain locations or workflows.

Claude is a generative AI system that can draft text, summarize documents, and help answer questions. For large employers, the approval process typically determines which data can be entered, what outputs can be relied on, and how usage is monitored under internal risk and compliance requirements.

The report characterizes the move as an internal tool-access decision, rather than a public product announcement. It does not, in the information available here, specify whether JPMorgan stopped using Anthropic’s technology entirely in Hong Kong, or whether it only restricted Claude’s availability to certain roles, teams, or use cases.

JPMorgan’s decision appears to fit a broader pattern in finance, where banks have increasingly moved to control AI deployment through governance, approved-tool lists, and location- or policy-based constraints. Regulators and internal risk teams have pushed firms to demonstrate that AI use does not compromise confidential information, trading or client data, or regulated decision-making.

Still, important details remain unclear from the published report alone. It is not stated how the restriction was implemented, when it took effect, whether other AI tools remain available to the same staff, or whether the policy change applies to both consumer-facing and internal “enterprise” versions of Claude.

The bank also did not disclose, in the information provided with the report, what specific standard it used to decide that Claude should no longer be approved in Hong Kong. Common reasons that banks cite for narrowing access include data handling concerns, model change management, auditability, and ensuring outputs are subject to appropriate human review.

For market watchers, the practical implication is that AI adoption in banking is not a simple yes-or-no rollout. Even when an AI assistant is approved in one region or business unit, banks may still restrict it elsewhere as risk assessments evolve or as internal controls are updated.

What to watch next is whether JPMorgan clarifies the scope of the restriction in Hong Kong, whether it replaces Claude with another approved tool, or whether it issues a broader statement about AI governance and approved model usage across regions. Without that, the report suggests a tightening of AI tool availability rather than a company-wide retreat from AI.