Anthropic is increasing enforcement measures after reports revealed Chinese companies successfully bypassed the startup's service ban to access its Claude large language models. Investigations by the Financial Times indicate that major firms, including Ant Financial and ByteDance, utilized various workarounds to circumvent restrictions, such as using overseas subsidiaries, virtual private networks (VPNs), and third-party cloud infrastructure like Microsoft Azure. These activities, while not violating US or Chinese law, breach Anthropic's terms of service, which explicitly forbid access by Chinese firms and entities under their control.

To counter these efforts, Anthropic is actively monitoring user accounts for suspicious indicators, including non-matching computer time zones and the use of relay services designed to route requests through foreign accounts. The company stated that it maintains a strict policy prohibiting Claude in unsupported regions and represents the only frontier AI firm that specifically restricts sales to PRC-controlled entities. Anthropic is now deploying continuous detection systems to identify and terminate accounts violating these policies.

This enforcement drive follows previous incidents of unauthorized access. In mid-June, Anthropic identified a large-scale distillation campaign involving thousands of fake accounts linked to Alibaba’s Qwen AI lab, which generated millions of interactions with Claude. To combat such threats, the company is reportedly implementing more rigorous verification protocols, potentially requiring government-issued IDs for suspicious accounts to ensure compliance with its international usage restrictions.

Anthropic is increasing enforcement measures after reports revealed Chinese companies successfully bypassed the startup's service ban to access its Claude large language models. Investigations by the Financial Times indicate that major firms, including Ant Financial and ByteDance, utilized various workarounds to circumvent restrictions, such as using overseas subsidiaries, virtual private networks (VPNs), and third-party cloud infrastructure like Microsoft Azure. These activities, while not violating US or Chinese law, breach Anthropic's terms of service, which explicitly forbid access by Chinese firms and entities under their control.

To counter these efforts, Anthropic is actively monitoring user accounts for suspicious indicators, including non-matching computer time zones and the use of relay services designed to route requests through foreign accounts. The company stated that it maintains a strict policy prohibiting Claude in unsupported regions and represents the only frontier AI firm that specifically restricts sales to PRC-controlled entities. Anthropic is now deploying continuous detection systems to identify and terminate accounts violating these policies.

This enforcement drive follows previous incidents of unauthorized access. In mid-June, Anthropic identified a large-scale distillation campaign involving thousands of fake accounts linked to Alibaba’s Qwen AI lab, which generated millions of interactions with Claude. To combat such threats, the company is reportedly implementing more rigorous verification protocols, potentially requiring government-issued IDs for suspicious accounts to ensure compliance with its international usage restrictions.