China's government is warning citizens against using US AI models through unofficial channels citing security risks, while US consumers increasingly turn to cheaper Chinese alternatives—signaling a deepening AI-sector divide between the two countries.

What happened: China's Ministry of State Security warned that users accessing US-based AI models like Anthropic's Claude through third-party marketplaces risk exposure to security threats, inadequate encryption, and data retention. Simultaneously, US consumers are adopting cheaper Chinese models such as DeepSeek V4 Pro, Alibaba's Qwen 3.6, and GLM 5.1, which offer lower inference costs than OpenAI and Anthropic.

Why it matters: The move reflects an escalating tit-for-tat dynamic—China is discouraging access to US AI by citing security concerns, mirroring US government restrictions on Chinese chip exports and AI hardware. Research from May 2026 documented a thriving ecosystem of API reseller services on Taobao, GitHub, and Telegram offering Claude access at as little as a tenth of the official price, often using bulk-registered accounts and stolen credentials. Both countries view foreign-user adoption as a strategic win, as more queries provide training data to improve their own AI models.

What to watch: Chinese authorities have signaled future restrictions on US AI access through advisories backed by state agency actions—the article notes a parallel in how China has discouraged but not blanket-banned Nvidia GPU imports for AI workloads. The US government may similarly escalate enforcement against citizens using foreign AI models as the competitive stakes for reaching artificial general intelligence (AGI) intensify.