Crypto exchange OKX has stopped its Hong Kong employees from using Anthropic's Claude, blockchain media outlet U.Today reported on Aug. 20 local time. The step was introduced earlier this month after the company's business account was temporarily suspended, and it also applied to employees passing through China.

The core issue is regional access restrictions. Anthropic policy bans access to Claude from regions it does not support. The current list of supported regions does not include Hong Kong or mainland China. OKX's business account was later restored, but access restrictions for staff in Hong Kong and those transiting China remained in place.

As a result, those employees' AI requests are being handled by other large language models. Staff had previously used both Claude and OpenAI's ChatGPT, but work is now running mainly on alternative models.

The move has drawn additional attention because it intersects with OKX's internal operating approach. OKX has pushed employees to use AI and has reflected AI proficiency in performance evaluations. The company also spends $6 million to $8 million a month on several major large language model providers. That means use of a specific model has been blocked even as AI is deeply embedded in daily work systems.

The issue is also seen as an example of U.S.-China AI tensions directly affecting global corporate operations. Anthropic policy blocks access from unsupported regions, and companies must adjust the scope of internal tool use to match that standard. For global companies with frequent cross-border travel, access can vary by region even when using the same account.

Similar cases have already emerged at other companies. Goldman Sachs faced a similar problem earlier this year. In April, employees in Hong Kong were unable to access Claude, and the restriction also applied to employees from other countries. Claude was reported to have been used mainly by software engineers at the time.

By contrast, Chinese companies are moving to impose their own blocks. Alibaba banned employees from using Claude Code in July. Alibaba classified Anthropic's coding assistant tool as high-risk software. Some companies had tried to access it via overseas subsidiaries, but that has become more difficult as restrictions have recently tightened.

Against this backdrop, the OKX case shows that even companies that have integrated AI tools into core work must frequently change their usage strategies as regional policies shift. As long as Hong Kong and mainland China remain excluded from supported regions, global crypto companies' AI operating systems are increasingly likely to face the same constraints.