JPMorgan Pulls the Plug on Anthropic for Hong Kong Staff
Wall Street Tightens Its Grip on AI Access
JPMorgan Chase has cut off its Hong Kong employees from Anthropic’s Claude AI models, joining a growing list of Wall Street banks that are clamping down on how staff use advanced artificial intelligence outside the United States. Employees in the Asian financial hub can no longer pull up Claude from the bank’s internal drop-down menu of approved large language models, according to people familiar with the move.
Related: OpenAI’s Spending Surged to $34bn in 2025 as IPO Looms
A Licensing Clause Drives the Decision
Rather than acting on a technical glitch or a security breach, JPMorgan made this call based on the fine print. One person close to the situation explained that the bank zeroed in on the specific wording of Anthropic’s usage terms, which carve out restrictions for Greater China. Because Anthropic’s licensing agreement excludes that region, JPMorgan chose to lock Hong Kong out entirely rather than risk a violation.
Goldman Sachs Got There First
JPMorgan isn’t blazing a new trail here. Earlier this year, Goldman Sachs made the same move, barring its Hong Kong bankers from tapping into Anthropic’s tools. Goldman likewise took a strict reading of Anthropic’s terms of use, which lump Hong Kong in with mainland China despite the territory’s separate legal traditions. Now, with two major banks aligned, a clear pattern is emerging across the industry.
Why Hong Kong Sits in a Gray Zone
To understand the stakes, consider Hong Kong’s unusual position. Mainland China bans Western AI models like OpenAI’s ChatGPT and Anthropic’s Claude outright, walling them off behind the so-called Great Firewall. Hong Kong, by contrast, has long operated free of Beijing’s censorship. As a result, the only barriers facing the city’s professionals come from the US AI companies themselves, not from local regulators.
Still, the workarounds have limits. While users can’t reach these models directly from Hong Kong, large international firms have historically sidestepped the geographic blocks by negotiating global contracts and routing their activity through servers outside China.
Related: European companies double down on China manufacturing despite EU de-risking push • Huawei unveils new smartphone chips this fall as rivalry with Nvidia and Apple heats up
The Distillation Fear Looms Large
So why are American AI labs so cautious? Much of their wariness traces back to “distillation” — the risk that local actors could effectively reverse-engineer a powerful model by feeding it enormous volumes of queries and learning from its outputs. Consequently, US AI groups guard access tightly, especially anywhere near the Chinese market.
That anxiety has already triggered dramatic action elsewhere. Just last week, Anthropic suspended all access to its cutting-edge Fable model after the US government urged the company to restrict foreign citizens, citing national security. Meanwhile, officials and financial executives have flagged concerns that Anthropic’s Mythos model could potentially exploit cybersecurity weaknesses across the global financial system.
Related: Nvidia earnings call drama: Will Jensen Huang talk ‘Trump’ and China chips after Xi summit?
What It Means for Hong Kong’s Comeback
Ultimately, this clampdown threatens more than convenience. As firms worldwide race to adopt AI for coding and analysis, cutting Hong Kong off from the most capable models risks dulling the city’s edge just as it works to reclaim its status as a premier international financial center.
For now, both sides are staying tight-lipped. JPMorgan declined to comment, and Anthropic did not immediately respond — though the company has previously noted that it never officially “supported” its Claude models in Hong Kong to begin with.
