Anthropic to close China loopholes

By the Editorial Team · Saturday, July 4, 2026

Anthropic is tightening restrictions to prevent Chinese companies from accessing its AI models by closing loopholes that allowed firms like Ant Financial and ByteDance to circumvent bans through overseas subsidiaries and cloud providers. This crackdown highlights the challenges U.S. AI firms encounter in enforcing usage policies amid strong demand for advanced AI tools in China and existing legal but unauthorized access methods.

Anthropic is intensifying measures to block unauthorized use of its artificial intelligence models by Chinese companies, closing loopholes that have allowed access despite strict restrictions. Sources familiar with the situation revealed that several Chinese firms, including Ant Financial, leveraged workarounds involving cloud providers and overseas subsidiaries to circumvent Anthropic’s prohibitions.

Employees at Ant Financial were reportedly given corporate accounts for Anthropic’s Claude Code, which were accessed via the company’s intranet connected to its Singapore-based entity. Meanwhile, ByteDance, the parent company of TikTok, does not facilitate direct access to Claude but introduced a policy allowing engineers to subscribe individually and claim expenses, according to five ByteDance employees. These engineers use virtual private networks (VPNs) to connect from mainland China.

Such practices do not violate U.S. or Chinese laws but contravene Anthropic’s terms of service. The company explicitly bars Chinese firms and their foreign subsidiaries from utilizing its models. Anthropic enforces one of the strictest bans among U.S. AI companies, requiring detailed user verification and prohibiting payments from Chinese financial institutions.

In contrast, Chinese users face fewer barriers accessing OpenAI’s tools via VPNs, as OpenAI does not implement the same level of user verification as Anthropic’s Claude platform. This disparity highlights the ongoing appeal and utility of leading American AI technologies to Chinese engineers, despite increasing regulatory hurdles and advancements in domestic AI models.

Anthropic’s coding tools are especially popular with Chinese software developers and AI startups because their outputs can be used for “distillation,” a process where smaller models are trained to replicate the capabilities of more advanced ones.

Efforts to plug these access gaps have proven challenging. Users have exploited pathways such as Microsoft’s Azure cloud infrastructure, connecting through foreign subsidiaries. Microsoft acknowledged that it provided application programming interface (API) access to Chinese companies via Singapore-based entities, allowing engineering teams in mainland China to use Claude within their corporate networks. The technology company stated that Anthropic monitors usage and enforces its terms with Microsoft’s assistance.

Requests for comment from Ant Financial and ByteDance went unanswered. The situation underscores the complexities U.S. AI companies face in enforcing usage restrictions amid a global market where demand for sophisticated AI tools remains high.

This article was reported by synthesizing coverage from multiple independent news outlets to present a balanced, unbiased account. Read about our editorial standards. Spotted an error? Let us know.