Goldman Sachs and OKX have restricted employee access to Claude: AI geopolitics is invading corporate finance.
The largest players in the financial sector — investment bank Goldman Sachs and crypto exchange OKX — have imposed a ban on the use of Anthropic's AI model Claude for their employees in Hong Kong. The restrictions were introduced a few months apart, but both companies faced the same problem: regional access rules.
Anthropic, the developer of one of the most advanced language models, does not provide access to Claude in Hong Kong and mainland China. This has created serious inconveniences for global corporations whose offices and employees operate in this region.
OKX: From Blocking to Resource Reallocation
Crypto exchange OKX encountered the problem after a temporary freeze of its corporate account at Anthropic. After restoring access, the company decided to redirect requests from employees in Hong Kong to alternative AI models. Notably, OKX does not plan to help workers bypass the block via VPN during business trips across the region.
The situation looks especially telling against the backdrop of active AI integration into the exchange's operational processes. According to OKX CEO Star Xu, the company spends between $6 and $8 million monthly on major language model providers. Moreover, the exchange encourages daily AI use and takes this factor into account when evaluating employee performance.
Goldman Sachs: Contractual Restrictions
Goldman Sachs closed access to Claude for employees in Hong Kong as early as early 2026. The bank discovered that its corporate agreement with Anthropic does not extend to this region. The restriction affected not only local workers but also foreign specialists traveling to Hong Kong on business trips. Claude was mainly used by software engineers.
Unlike OKX, the reason for the restriction at Goldman Sachs lies in the terms of the contract, not a temporary account block. At the same time, the bank is not abandoning cooperation with Anthropic — Chief Information Officer Marco Argenti has been working jointly with the company's engineers for several months to create AI solutions for accounting for trading operations and client verification.
Hong Kong authorities actively support the adoption of artificial intelligence in finance, but restrictions from developers themselves, including Anthropic, hinder this. Both companies faced the same geographic barriers, pointing to a global trend.
Washington and Beijing are actively competing for leadership in AI. Chinese models are increasingly catching up with American ones.
Because of this rivalry, American companies decide for themselves who can use their technologies and where. Neither company specified whether the restrictions extend to clients and partners from mainland China. This issue becomes especially relevant ahead of the September US-China talks on AI.
My analysis: This situation is a clear marker of the fragmentation of the global technological landscape. For the crypto industry, which has historically been built on the principles of decentralization and free access, dependence on regional restrictions from American AI providers creates new operational risks. Exchanges and financial institutions will have to diversify their AI stacks, including Asian alternatives, which in the long term could accelerate the development of local models.