JPMorgan Chase restricts Anthropic AI access for its Hong Kong staff, citing internal controls

JPMorgan Chase has limited the use of Anthropic’s AI models by its staff in Hong Kong, according to a Yahoo Finance report published June 18. The move reflects a broader pattern among major banks to tighten internal controls around generative AI tools, particularly in regions where regulatory expectations, data handling, and model governance are under heightened scrutiny.

The report says JPMorgan Chase stopped Hong Kong-based employees from accessing Anthropic’s models. It describes the change as following a similar decision by Goldman Sachs, another large Wall Street firm that has publicly or at least reported internal restrictions around access to certain external AI services.

While the report frames the action as an “access cut-off,” it does not provide additional specifics on which Anthropic models were affected, whether the bank paused all Anthropic usage or only direct employee access, or what technical or policy mechanism JPMorgan used to enforce the restriction. It also does not state whether JPMorgan replaced Anthropic access with an alternative internal AI workflow or with another vendor’s tools.

JPMorgan’s step sits in the context of how banks are experimenting with generative AI for tasks such as document review, client support, research assistance, and software development, while simultaneously confronting concerns about data privacy, confidentiality, and the reliability of model outputs. In practice, many banks are moving toward “approved” systems, centralized gateways, and stricter permissions rather than leaving access open to large numbers of users.

Because Yahoo Finance’s report is not accompanied here by a JPMorgan statement or a regulatory filing, key details remain unclear. The bank did not spell out whether the restriction was driven by compliance reviews, risk management concerns, contractual changes with the vendor, or broader internal governance policies for AI usage in Hong Kong.

For market participants, The announcement is less about a specific vendor and more about how quickly compliance and risk teams are shaping what employees can do with AI. Even when banks adopt AI experimentally, they can still reduce access if controls do not meet internal standards or if regulators or auditors raise questions about how data and outputs are handled.

Going forward, investors and customers are likely to watch whether JPMorgan discloses more about its AI governance framework, whether other offices or business lines follow suit, and whether the bank publicly confirms whether it will rely on Anthropic models through more controlled channels or shift to different systems for permitted use cases.