OKX tightens AI controls after Anthropic account suspension

The suspension, which occurred in early August, prompted an internal review by the exchange. OKX officials acknowledged that previous employee activity may have veered outside of Anthropic’s geographical access rules. While the enterprise account has been restored, staff in these restricted regions are now being funneled toward alternative artificial intelligence services to prevent further policy breaches.

This incident underscores the challenges of managing global software contracts in regions where specific providers limit commercial access. Anthropic currently excludes both Hong Kong and mainland China from its list of supported territories for commercial API usage. Similar measures have been observed at major financial institutions, including JPMorgan and Goldman Sachs, which have likewise moved to restrict Claude access for staff based in or traveling through these jurisdictions.

Beyond these access hurdles, the situation reveals the massive scale of OKX’s investment in automation. The exchange disclosed a monthly expenditure of $6 million to $8 million on AI providers, totaling nearly $100 million annually. This spending supports a strategy that integrates AI directly into developer workflows and employee performance evaluations. Despite the internal disruption, OKX confirmed that the suspension did not impact its exchange operations, client assets, or the broader trading platform.