Goldman blocks Claude use for Hong Kong bankers

Goldman Sachs has blocked its bankers in Hong Kong from using Anthropic’s Claude on an internal AI platform in recent weeks after a review of its contract with the AI startup.

The bank took a strict reading of the agreement after consulting Anthropic and concluded staff in Hong Kong should not use any Anthropic products, while ChatGPT and Gemini remained available.

US-built models such as ChatGPT and Claude are barred in mainland China, while Hong Kong has largely stayed outside those controls.

Goldman said earlier in February it was working with Anthropic on AI agents for internal tasks.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

Anthropic ties AI access to ownership as well as location

Anthropic bars access for organizations that are more than 50% owned by companies based in unsupported regions such as China, even when those groups operate elsewhere 1.
The company said the rule aims to reduce national security risk. Firms controlled from authoritarian regions may face legal pressure to share data or help intelligence services 2.
In practice, access can depend on who controls a business, not only where staff are based.
Goldman Sachs’ move fits a strict reading of those terms and may help it avoid breaking the contract, rather than follow a simple location ban 3.

Geopolitics makes cross-border AI rollouts harder to manage

For multinational companies, deploying AI across markets can get tangled in contract terms, export controls, and security rules.
The effect reaches beyond finance. It also clouds the outlook for products tied to Chinese owners, including ByteDance-backed Trae, an AI coding editor that uses Anthropic’s Claude, where some users have asked for refunds 3.
The limits may give local rivals more room to win customers.
Chinese AI company Z.ai, formerly called Zhipu AI, launched special offers for Claude application programming interface (API) users looking for other options 3.