Detailed Analysis
Anthropic's Claude, along with other American AI systems such as OpenAI's ChatGPT, has become widely used inside China despite the fact that these services are officially blocked and not licensed to operate in the country. According to reporting from NBC News, Chinese users are circumventing the Great Firewall through VPNs, overseas phone numbers, and third-party payment workarounds to access tools that Beijing has not approved for domestic distribution. This underground demand reflects a broader pattern in which sophisticated Chinese consumers, developers, and professionals perceive American frontier models as qualitatively superior to domestic alternatives for certain tasks, even as homegrown competitors like DeepSeek, Alibaba's Qwen, and Moonshot's Kimi have made rapid technical progress and dominate the legal, above-board market.
The dynamic underscores a paradox at the heart of the U.S.-China AI competition: while the two nations are locked in an intensifying rivalry over chip export controls, compute infrastructure, and model capability benchmarks, the actual on-the-ground behavior of Chinese users suggests that American models still carry a premium in perceived quality, reliability, or utility—particularly for tasks like coding, research, writing, and creative work where Claude has built a strong reputation. This matters because it reveals that government-level restrictions and geopolitical friction do not necessarily suppress grassroots demand; instead, they push usage into gray-market channels that are harder to monitor, monetize, or regulate. For companies like Anthropic, this creates a strange situation where a significant user base exists in a market the company cannot officially serve, generating no direct revenue while still shaping the model's global reputation and usage patterns.
This phenomenon also complicates the narrative that China's AI ecosystem is fully self-sufficient or indifferent to Western models. Despite state-backed investment in domestic champions and rhetoric about technological sovereignty, the persistence of black-market access to Claude and ChatGPT suggests gaps in capability, trust, or specific feature sets that domestic alternatives haven't fully closed. It also raises thorny questions for Anthropic and its peers around content moderation, data security, and compliance, since usage originating from China occurs entirely outside any contractual or regulatory framework the company has established—meaning Anthropic has no visibility into how its models are being used, by whom, or for what purposes within Chinese borders.
More broadly, this story fits into a growing trend of AI diffusion outpacing the ability of any single government to fully contain it. Just as encrypted messaging apps and VPNs have long found ways around censorship regimes, AI chatbots are proving similarly resistant to hard borders, especially as access can be achieved through browser extensions, third-party interfaces, and API wrappers rather than official app-store distribution. For Anthropic, this presents a strategic dilemma: the company has taken a relatively cautious, safety-focused stance on international expansion, particularly regarding adversarial nations, yet unofficial demand suggests its brand and technology are penetrating markets regardless of formal policy. This tension between deliberate corporate caution and organic global demand is likely to remain a recurring theme as frontier AI models continue to diffuse across geopolitical fault lines faster than regulatory frameworks can adapt.
Read original article →