Detailed Analysis
Goldman Sachs has blocked its Hong Kong-based employees — including software engineers and investment bankers — from accessing Anthropic's Claude AI models through the firm's internal platform. The restriction, which has been in place for several weeks as of late April 2026, is geographically enforced and extends to employees who are merely traveling to Hong Kong, not just those permanently stationed there. According to sources with direct knowledge of the matter, the block stems from Goldman Sachs' interpretation of its contractual agreement with Anthropic, under which the bank determined that Hong Kong-based usage falls outside the permitted scope of the license. Prior to the restriction, affected staff had been using Claude for productivity-intensive tasks including coding assistance and financial modeling.
The practical impact of the decision is notable because it creates an uneven internal technology landscape within one of the world's largest financial institutions. Hong Kong employees still retain access to competing AI tools — most prominently OpenAI's ChatGPT — meaning the restriction is product-specific rather than a blanket prohibition on generative AI use. This asymmetry underscores that the block is rooted in contractual and geographic licensing constraints rather than a broader firm-wide policy concern about AI safety or data privacy. It also highlights the operational friction that can arise when enterprise AI deployments encounter the realities of multinational corporate structures and territorially defined service agreements.
The development reflects a broader challenge facing AI companies as they scale enterprise adoption globally: product availability and licensing frameworks do not always map cleanly onto the geographic complexity of large multinational clients. Anthropic's Claude lacks official support in Hong Kong, a gap that Goldman's legal and compliance teams appear to have treated as determinative when interpreting their contract. This stands in contrast to OpenAI, whose products remain accessible to the same employees, suggesting that OpenAI has either secured broader geographic licensing agreements with Goldman or has official product support in the Hong Kong market that Anthropic has not yet established.
For Anthropic, the episode illustrates both a commercial limitation and a strategic opportunity. Losing access within a prestigious financial services client — even on a regional basis — carries reputational and revenue implications, particularly as Wall Street firms have become key enterprise customers for frontier AI models. The situation signals that Anthropic may need to accelerate its geographic coverage and clarify licensing terms to prevent similar restrictions at other large institutions operating across multiple jurisdictions. As the enterprise AI market matures, the ability to offer consistent, globally supported access is likely to become a significant differentiator in competitive contract negotiations between major AI providers and multinational corporations.
Read original article →