OKX has restricted employee access to Claude, the AI model built by Anthropic, for staff working out of its Hong Kong offices. The move was reported by CryptoBriefing and crypto.news, both citing the exchange's internal handling of AI tools used by its workforce.
Alongside the restriction, both outlets reported that OKX's total monthly spending on AI models has reached between $6 million and $8 million. That figure covers the range of AI systems the exchange relies on for tasks spanning customer support, trading infrastructure, compliance monitoring and internal software development.
The scale of that spending illustrates how central AI tools have become to day-to-day operations at major crypto exchanges. Firms like OKX increasingly depend on large language models not just for experimental projects but for core business functions, from automating support tickets to assisting engineers with code review.
Why Claude access has been curtailed specifically for Hong Kong employees was not detailed in the reports. Global technology firms have faced a patchwork of restrictions on advanced AI models tied to jurisdiction, driven by export controls, data residency requirements and national security concerns tied to the broader US-China technology rivalry. Hong Kong, as a special administrative region with close ties to mainland China, often sits at the center of these compliance questions for Western AI providers.
Anthropic, like other leading AI developers, has faced scrutiny over where and how its models can be accessed given ongoing geopolitical tensions around advanced computing and AI exports. Companies operating across multiple jurisdictions, including exchanges like OKX with a presence in Hong Kong, have had to navigate these evolving rules when deciding which AI tools employees in different offices can use.
OKX operates as one of the larger global cryptocurrency exchanges, with a footprint spanning trading, custody and a range of financial products. Its reliance on AI models sits within a broader industry trend of exchanges building AI into risk management, market surveillance and customer-facing services. The reported monthly spend places OKX among the more visible corporate consumers of commercial AI infrastructure within the crypto sector.
Neither report detailed which alternative AI models, if any, Hong Kong staff have been directed to use instead of Claude. It also remains unclear whether the restriction reflects an internal OKX compliance decision, a policy set by Anthropic, or a response to broader regulatory guidance affecting the region.
Market Impact
The disclosure of OKX's AI spending offers a rare window into how much major exchanges now allocate toward AI infrastructure as a core operating cost. A monthly outlay of $6 million to $8 million suggests AI has moved from a discretionary innovation budget line to a recurring, significant expense for large trading platforms.
For the broader crypto industry, the restriction on Claude access in Hong Kong may signal how geopolitical and regulatory considerations are shaping which AI tools exchanges can deploy across different jurisdictions. Other exchanges with multi-regional operations could face similar decisions about which AI providers are usable in which offices, particularly where compliance and data-handling rules diverge by region.
The reported restriction and spending figures highlight how intertwined AI infrastructure has become with exchange operations, and how jurisdictional compliance questions can shape even routine employee tool access.
Frequently Asked Questions
What did OKX restrict for its Hong Kong employees?
According to CryptoBriefing and crypto.news, OKX limited access to Anthropic's Claude AI model for staff working in its Hong Kong offices.
How much does OKX reportedly spend on AI models each month?
The reports put OKX's monthly AI model spending at between $6 million and $8 million across its operations.
Why was Claude access limited specifically in Hong Kong?
The reports did not specify a reason. Restrictions on AI models by jurisdiction often relate to export controls, data residency rules or regulatory considerations tied to a region.
Does this affect OKX's use of AI in other regions?
The reports focused on Hong Kong-based staff and did not indicate similar restrictions in OKX's other offices.