JPMorgan Cuts Anthropic AI Access for Hong Kong Staff

JPMorgan Chase removed Anthropic's Claude AI from its approved tools list for Hong Kong staff, mirroring a move Goldman Sachs made in April 2026.

JPMorgan removed Anthropic's Claude models from its approved LLM list for Hong Kong employees over licensing and data security concerns.
Goldman Sachs took an identical step in April, with Hong Kong staff barred from Claude while other tools such as Gemini and ChatGPT remain available.
The actions follow Anthropic's explicit policy excluding Greater China from supported markets and accelerating U.S. government pressure on AI model exports.

Lead

JPMorgan Chase has barred its Hong Kong staff from accessing Anthropic's Claude AI models, removing the technology from an internal drop-down list of approved large language models available to employees in the city, the Financial Times first reported on June 18, 2026. The move makes JPMorgan the second major Wall Street institution within two months to restrict Hong Kong AI data exposure to Anthropic's tools, following a near-identical decision by Goldman Sachs in late April.

What Happened

JPMorgan employees in Hong Kong can no longer select Claude from the bank's internal platform of approved AI tools. The restriction stems directly from the wording of Anthropic's licensing agreement with the bank: Anthropic's usage terms explicitly exclude Greater China — a designation that encompasses Hong Kong — from its list of supported markets. Faced with ambiguity about whether deploying Claude for staff in the territory would breach those geographic provisions, JPMorgan removed the model from the approved list rather than risk a compliance violation.

Other widely used AI models, including Google's Gemini and OpenAI's ChatGPT, remain available to JPMorgan staff in Hong Kong under separate licensing arrangements that carry no equivalent geographic restrictions.

Goldman Led the Way

Goldman Sachs preceded JPMorgan in making this call. In late April, Goldman removed Claude from its internal AI platform for Hong Kong-based employees after adopting a strict reading of its own contract with Anthropic following consultations with the company. The restriction applied regardless of home office: staff visiting Hong Kong from other offices found themselves blocked from Claude while in the city. At Goldman, Claude had been predominantly used by software engineers for code assistance and productivity tasks.

Anthropic Cybersecurity Policy and U.S. Export Pressure

The bank AI policy decisions are unfolding against a backdrop of escalating U.S. scrutiny of AI model exports. U.S. Commerce Secretary Howard Lutnick directed Anthropic to suspend worldwide exports of its newer Mythos and Fable AI models to all foreign nationals, citing concerns that the technology could be accessed by military intelligence services in China, Russia, and other countries identified as security risks.

Anthropic has articulated a clear Anthropic cybersecurity posture of its own. In a June 11 policy proposal, the company urged governments to establish independent testing requirements, cybersecurity standards, and enforcement mechanisms for frontier AI systems, arguing that the most advanced models carry biological, operational, and cybersecurity risks that require structured oversight. Separately, Anthropic updated its regional sales restrictions policy to formally exclude Greater China from its supported API markets — a move that directly triggered compliance reviews at both banks.

Geopolitical Dimension

Hong Kong's status as a Special Administrative Region means companies operating there are subject to mainland Chinese law, including data security and national security statutes that can compel the sharing of information with state authorities. That legal environment creates a structural incompatibility with Anthropic's terms of service, which prohibit use in jurisdictions where data could be compelled by adversarial state actors.

The decisions by JPMorgan and Goldman reflect a broader reorientation in how JPMorgan Anthropic access and analogous enterprise AI arrangements are being structured across financial institutions with global footprints. Banks in Hong Kong are effectively navigating a binary trade-off: full compliance with AI vendor licensing terms, or full tool access for Asia-based workforces. That friction is reshaping enterprise AI deployment strategies across the region at a meaningful pace.

Outlook

With Anthropic's licensing restrictions now enforced by at least two of the largest U.S. financial institutions, the precedent is likely to prompt broader reviews at other banks and asset managers with significant Hong Kong operations. The underlying tension — between frontier AI vendors' geographic restrictions and the operational needs of globally distributed financial firms — is structural and will not resolve without either a regulatory accommodation in Hong Kong or a recalibration of Anthropic's licensing framework. As U.S.-China technology friction persists and Anthropic advances toward a potential public offering, its stance on Greater China will remain a closely watched variable for enterprise clients worldwide.