Goldman Sachs Pulls the Plug on Anthropic's AI in Hong Kong — While Doubling Down at Home

.

Claude Goes Dark for Goldman's Hong Kong Desk

Something changed for Goldman Sachs employees in Hong Kong a few weeks ago: access to Anthropic's Claude AI models simply stopped working. According to reporting by the Financial Times, the Wall Street banking giant has blocked its Hong Kong staff from using any Anthropic products — a decision that was not driven by regulation, but by a contractual interpretation.

Four sources familiar with the matter told the FT that Goldman reviewed its agreement with Anthropic and concluded, after direct consultation with the company, that Hong Kong employees should not have access. The restriction does not apply to other AI vendors Goldman works with, such as OpenAI.

A Gray Zone Turns Black and White

Hong Kong occupies a complicated spot in the global AI landscape. Mainland China prohibits US-built AI models like Claude and ChatGPT from operating freely — but Hong Kong has, up until now, largely remained outside those controls. In practice, usage limits in the region have been set by the US companies themselves, not by Beijing.

That calculus is shifting. Anthropic told the FT that Claude had never been officially "supported" in Hong Kong — a telling admission that the region was always in a gray zone rather than a formally approved market. The company declined to elaborate further.

This is consistent with a broader tightening Anthropic undertook in September 2025, when it updated its terms of service to ban access not just by geography, but by ownership structure. Any company more than 50 percent owned — directly or indirectly — by entities headquartered in restricted regions like China was cut off globally, regardless of where it operates. Hong Kong's post-2020 legal environment, including the National Security Law, plays a central role in how Anthropic categorizes the territory.

The Contradiction: Partner Abroad, Blocked at Home

What makes Goldman's Hong Kong decision striking is the contrast with what's happening at the bank's US offices. Just this past February, Goldman's Chief Information Officer Marco Argenti announced a deepening partnership with Anthropic — one of the most significant AI collaborations on Wall Street.

The bank had spent six months embedding Anthropic engineers directly inside its teams to co-develop autonomous AI agents aimed at automating core banking functions: trade reconciliation, transaction accounting, and client onboarding. Argenti described the tools as "digital co-workers" for complex, process-heavy jobs. Early tests reportedly showed client onboarding times cut by 30 percent and developer productivity up more than 20 percent.

In short: Anthropic's technology is being positioned as a cornerstone of Goldman's AI future — just not everywhere Goldman operates.

What This Signals for Wall Street's AI Race

The episode highlights the increasingly fragmented nature of global AI access. US AI companies are not just responding to market forces — they are making deliberate geopolitical decisions about who gets access to their most powerful tools.

Anthropic has been explicit about its reasoning. In its September 2025 policy update, the company argued that firms operating in or under the influence of authoritarian jurisdictions face legal requirements that can compel data-sharing with intelligence services — making it difficult for any US AI provider to safely serve those markets, regardless of the end user's intentions.

OpenAI made similar moves last year, tightening its own access policies to block usage from unsupported countries and territories. China, meanwhile, has built its own AI approval regime — one that currently includes only domestically developed models.

For Goldman, the Hong Kong ban appears to be less a rejection of Anthropic and more a careful compliance move — one that reflects how seriously large financial institutions now take AI vendor contracts and the geopolitical strings attached to them.

.

Sources

1. Reuters – Goldman Sachs bars Hong Kong bankers from Anthropic AI use: https://www.reuters.com/world/china/goldman-sachs-bars-hong-kong-bankers-anthropic-ai-use-ft-reports-2026-04-29/
2. CNBC – Goldman Sachs taps Anthropic's Claude to automate accounting, compliance roles (February 6, 2026): https://www.cnbc.com/2026/02/06/anthropic-goldman-sachs-ai-model-accounting.html
3. South China Morning Post – Anthropic blocks Chinese firms' subsidiaries worldwide from AI access (September 5, 2025): https://www.scmp.com/tech/tech-war/article/3324504/tech-war-us-start-anthropic-blocks-chinese-firms-subsidiaries-worldwide-ai-access
4. Anthropic – Updating restrictions of sales to unsupported regions: https://www.anthropic.com/news/updating-restrictions-of-sales-to-unsupported-regions
5. South China Morning Post – Black market workarounds for Claude scale up as Anthropic tightens ID checks (April 2026): https://www.scmp.com/tech/article/3350321/black-market-workarounds-claude-scale-anthropic-tightens-id-checks

.