Goldman Sachs has officially restricted its Hong Kong-based bankers from using Claude AI, the advanced model developed by Anthropic.
While the bank has aggressively integrated AI into its global operations, this targeted blockade highlights the complex “legal gray area” that Hong Kong currently occupies in the escalating AI trade war between the U.S. and China.
1. The Reason: A “Strict” Contractual Interpretation
According to internal sources and reporting from the Financial Times, the ban is not a rejection of AI technology but a matter of compliance and risk management.
- Unsupported Market: Anthropic’s official website does not list Hong Kong as a supported jurisdiction for its API or consumer products.
- The Review: Following a consultation with Anthropic, Goldman Sachs conducted a “strict interpretation” of its enterprise contract. The bank concluded that allowing staff in Hong Kong to use Claude—even via internal “walled” platforms—could violate the terms of their agreement.
- Selective Ban: Notably, ChatGPT (OpenAI) and Gemini (Google) remain accessible to Goldman bankers in Hong Kong. This suggests that OpenAI and Google have different contractual or geographical permissions than Anthropic regarding the territory.
2. Geopolitics & “Technology Leakage”
The restriction comes just weeks before a planned mid-May 2026 summit between Presidents Donald Trump and Xi Jinping, where AI security is expected to be a primary focus.
- The “Distillation” Threat: U.S. AI labs are increasingly wary that intensive use of their models by firms operating near mainland China could lead to “distillation”—a process where local actors use a foreign AI’s outputs to train competing domestic models.
- Data Sovereignty: Global banks are under pressure to ensure that sensitive financial data processed by AI doesn’t cross into jurisdictions where local intelligence laws could force its disclosure.
- Hong Kong’s Unique Status: While mainland China is behind the “Great Firewall” (blocking all Western AI), Hong Kong has traditionally been a “free port” for data. However, U.S. companies are now voluntarily imposing their own limits to avoid running afoul of tightening U.S. export controls.
3. Impact on the Banking Workforce
The ban on Claude is particularly significant because of the model’s reputation for high-end financial modeling and coding.
- Productivity Gap: Bankers who relied on Claude for complex spreadsheet automation or python-based data analysis may now face a “speed disadvantage” compared to their colleagues in New York or London who still have full access.
- Regulatory Scrutiny: The Hong Kong Monetary Authority (HKMA) has reportedly contacted major lenders to gather information on their use of advanced models like Claude Mythos, reminding them to update their risk assessments for “agentic AI” that can execute tasks autonomously.
- The Ripple Effect: Analysts expect other Wall Street firms like Morgan Stanley and JPMorgan to review their own regional AI deployments to ensure they aren’t in breach of similar “territory-specific” clauses in their tech contracts.