Cryptocurrency exchange OKX has restricted employees in Hong Kong and staff traveling through mainland China from using Anthropic’s Claude model after its enterprise account was briefly suspended in early August. The suspension was later resolved, but OKX said some earlier usage may not have complied with Anthropic’s regional access policies covering Greater China operations.
According to internal messages reviewed by Bloomberg, OKX will redirect affected requests to other artificial intelligence models to avoid further breaches. The exchange has not disclosed how many employees were affected or which alternative AI providers will handle requests from restricted locations. Anthropic does not offer Claude access in Hong Kong or mainland China, forcing companies with global contracts to control where employees use the service.
The restriction highlights how deeply AI has become embedded in OKX operations. The company spends between $6 million and $8 million monthly across major AI providers, translating to an annualized run rate of roughly $72 million to $96 million. AI adoption is also reportedly part of employee performance evaluations as OKX expands internal model usage globally.
OKX is not alone in facing these geographic limitations. Goldman Sachs and JPMorgan have previously removed or restricted Claude access for Hong Kong employees after reviewing Anthropic’s licensing terms. The situation shows that AI-provider policies can now determine which software tools employees of a global crypto exchange can use depending on their physical location, making model availability an infrastructure, compliance and geopolitical consideration.
The internal AI restriction did not concern OKX’s exchange operations, customer assets or trading services, and no verified market reaction was linked to the event.
