Goldman Sachs cuts Anthropic AI access for Hong Kong bankers over data concerns

Goldman Sachs has cut Anthropic AI access for bankers in Hong Kong, a move that highlights how the bank’s AI rollout now depends as much on location and data rules as on productivity gains.

The restriction is not a companywide pullback. It is location-specific, and it blocks Goldman staff visiting Hong Kong from overseas as well. The Claude tools had been available through an internal AI platform, but bankers in the city no longer have access, a change that lands as global banks face sharper scrutiny of AI tools and data security. Bloomberg said the technology in Hong Kong had been used predominantly by software engineers, making the loss especially operational for teams that had folded it into day-to-day coding and workflow tasks.

For Goldman employees, the practical effect is immediate: the same firm can offer different AI capabilities depending on where a person is sitting. A banker in New York or London may still have access to approved tools, while a counterpart in Hong Kong may need alternate workflows for drafting, coding, research, or internal automation. That matters inside a business where speed, control, and consistency are part of the job, and where any delay can ripple into client work, trading support, or technology delivery.

The decision also shows Goldman is not stepping away from Anthropic. Chief information officer Marco Argenti said in February that Goldman was working with Anthropic on AI-powered agents designed to automate a widening range of internal functions. CNBC reported that embedded Anthropic engineers had spent six months at Goldman co-developing autonomous agents for trade accounting and client onboarding. Goldman Sachs Research has framed AI agents as a major productivity milestone, saying they can lift business productivity and expand the software market.

Anthropic has also been pushing deeper into finance. Claude for Financial Services launched in July 2025, then expanded in October with an Excel add-in, real-time market-data connectors, and pre-built agent skills for tasks such as discounted cash flow models and coverage reports. At the same time, Anthropic says its commercial API and Claude.ai are only available in supported countries and regions, underscoring that access has always been shaped by jurisdiction and compliance.

Hong Kong had already signaled where regulators stood. The Financial Services and Treasury Bureau issued a policy statement on responsible AI in financial markets in October 2024, stressing a risk-based approach and flagging cybersecurity, intellectual property, and data-privacy concerns. For Goldman, that makes Hong Kong a case study in the new AI workplace reality: the tools are powerful, but who gets them now depends on where the employee is, what the contract allows, and how much risk the bank is willing to tolerate.