JPMorgan Prohibits Hong Kong Employees from Accessing Anthropic Claude Model
JPMorgan has stopped employees at its Hong Kong branch from accessing the Claude AI model developed by Anthropic. Employees cannot select this tool from the approved list of large language models, following similar restrictions imposed by Goldman Sachs earlier this year.
The decision stems from restrictive language in Anthropic's terms of use that excludes the Greater China region (including Hong Kong) from the service scope. The bank interprets the licensing agreement strictly, making it impossible to circumvent geographical terms.
Wall Street banks are tightening AI access controls in Hong Kong, with pressure from compliance and data security departments. This has put pressure on trading and investment banking operations that rely on cutting-edge models to enhance coding efficiency. Capital and talent may flow to regions without such restrictions, weakening Hong Kong's competitiveness as an international financial center.
Source: Public Information
ABAB AI Insight
JPMorgan, as a leading global investment bank, has long integrated AI tools into trading, risk control, and coding processes, previously adopting various large models to enhance efficiency. The recent ban in Hong Kong continues its global compliance-first strategy, aligning with Goldman Sachs' actions, highlighting the bank's unified control over cross-border data risks.
On the capital front, the bank attempts to bypass restrictions through global contracts and overseas computing power, but the terms barriers force local teams to turn to compliant alternatives or reduce AI usage intensity. Resources are being redirected to the U.S. and friendly jurisdictions, while accelerating investments in local or open-source models to reduce dependency.
Similar to Goldman Sachs' earlier ban and adjustments by several multinational companies to global AI policies, the financial industry is undergoing a transformation dominated by geopolitical compliance. Hong Kong, as an Asian hub, faces disadvantages in tool availability.
Essentially, this reflects regulatory changes and a shift in pricing power, with mechanisms in place for U.S. companies to build technological barriers through export control-like terms to prevent model distillation risks, leading to a tilt of capital and innovation resources towards controlled ecosystems, further reshaping the global fintech industry chain.
ABAB News · Cognitive Law
Terms barriers are harder than firewalls; when compliance is prioritized, geography becomes the new national boundary.
AI Efficiency = Access × Computing Power; restricting one side permanently cedes competitiveness.
Technological openness accelerates diffusion, while geopolitical closure accelerates bifurcation, turning winner-takes-all into winner-takes-divided.