Anthropic, the U.S. artificial‑intelligence firm behind the Claude large‑language‑model, is intensifying efforts to prevent unauthorized access by Chinese entities after uncovering several workarounds. The Financial Times, citing sources, reported that firms such as Ant Financial accessed Claude through a Singapore‑registered subsidiary that provided corporate accounts to its employees, while ByteDance reimbursed engineers for personal Claude subscriptions obtained via virtual private networks. Additional loopholes involved the use of overseas subsidiaries and cloud infrastructure, including Microsoft’s Azure platform, to relay requests to Claude accounts located outside China. Anthropic’s terms of service expressly forbid Chinese companies and any foreign entities under their control from using its models; the identified practices breach those terms even though they do not contravene U.S. or Chinese law. To detect and shut down such access, Anthropic is monitoring indicators such as the time‑zone settings of user computers and targeting “transfer‑station” services that act as intermediaries for requests routed through foreign Claude accounts. The company says it will continue to block accounts that exhibit these indicators and to enforce its usage policies.