Anthropic is stepping up efforts to close loopholes that have allowed Chinese companies to access its AI tools despite one of the strictest bans among U.S. frontier AI companies. Ant Financial provided employees with corporate Claude accounts accessed through its intranet connected to a Singapore-based entity, while ByteDance introduced a reimbursement scheme allowing engineers to expense personal Claude subscriptions and access them via VPN. The workarounds breach Anthropic's terms of service but do not violate U.S. or Chinese law. Neither company responded to requests for comment.
Microsoft NASDAQ:MSFT sold API access to Chinese companies with Singapore entities, allowing mainland engineers to reach Claude through internal networks. Anthropic has also been targeting "transfer station" services that relay requests from mainland China through overseas Claude accounts, though larger Chinese AI groups avoid these because of concerns that operators store or resell prompts. Anthropic said it "explicitly prohibits accessing or facilitating access to Claude in unsupported regions, including China" and uses continuously evolving detection systems to identify and ban violating accounts.
Claude Code is particularly valued by Chinese engineers because its outputs can be used for distillation, a technique for training smaller models to mimic more capable ones.