State Department flags "serious concerns" as Chinese AI adoption spreads through US firms

A State Department spokesperson told CNBC that US corporate use of Chinese AI models "raises serious concerns," describing the models as "designed to advance Beijing's narratives, censor dissent and reflect CCP ideology and values." CNBC The spokesperson was not named. The remarks accompany reporting that US lawmakers are now weighing strategies to slow the growing adoption of Chinese models inside American companies. CNBC

The adoption pattern is no longer confined to obscure startups. Coinbase CEO Brian Armstrong disclosed that his company uses two Chinese models — GLM 5.2, built by Z.Ai, and Kimi 2.7, from Moonshot — a disclosure that CNBC reported alongside the State Department comments. CNBC Nikkei Asia separately reported a broader surge among American firms including Airbnb and Uber. Startup Lindy switched to DeepSeek in June specifically to curb rising inference costs, CNBC reported at the time. CNBC

Two mechanics explain why this has happened so quickly. First, cost: CNBC reported on July 7 that the share of tokens routed to Chinese models by US companies on the OpenRouter platform has risen significantly through 2026, tracking a broader shift toward efficiency as OpenAI and Anthropic pricing strains budgets. CNBC Second, distribution: because many leading Chinese models are released as open weights, US firms can simply download and self-host them, sidestepping API gatekeeping entirely. CNBC reported on June 30 that this ease of access is central to why Chinese developers are reaching US users at all. CNBC

Nikkei Asia's reporting adds a causal wrinkle: the surge correlates with Anthropic's suspension of its Mythos 5 and Claude Fable 5 models at the US government's request. Restricting the frontier domestically, in other words, appears to have pushed some demand toward Chinese alternatives rather than eliminating it. CNBC reported June 30 that the administration's broader clampdown on AI access could give Chinese labs time to close the capability gap with Anthropic and OpenAI. CNBC

Washington's policy architecture has been building toward this moment for a year. The White House's "America's AI Action Plan," published July 2025, recommended withholding federal AI funding from states with what it called burdensome regulations. A December 2025 presidential action directed evaluation of state AI laws under a national policy framework. In March 2026 the administration unveiled a national AI legislative framework, and on April 23 the White House issued NSTM-4, a national security telecommunications memorandum addressing model security protocols and what it terms ideological neutrality. Executive Order 14409, "Promoting Advanced Artificial Intelligence Innovation and Security," followed on June 2. White House CNBC reported June 17 that the US was separately pressing allies to restrict foreign access to leading American-developed models during a summit in France that concluded without Beijing's participation. CNBC

Export controls have tightened in parallel. Reuters reported May 31 that the US moved to halt Nvidia AI chip shipments to Chinese firms operating outside China, amid suggestions that top-tier chips were reaching subsidiaries of Chinese AI companies through indirect channels. Reuters In April, Reuters reported new US restrictions on investment in Chinese AI, semiconductor and quantum sectors, and separately that Beijing was preparing reciprocal curbs on US investment in Chinese tech firms. Reuters

Beijing's own posture toward its AI exports has shifted too. TIME reported July 7 that Chinese AI firms built global market share by offering models for free, and that Beijing is now weighing restrictions on overseas access to its most capable systems. TIME Reuters reported similarly on July 7 that Chinese authorities have opened talks with domestic companies aimed at preventing homegrown AI from being used overseas at all. Reuters Chinese regulators have also escalated their own security allegations: the Ministry of Industry and Information Technology announced that Anthropic's Claude Code contains a backdoor it calls a "serious threat." engadget

Security concerns are not confined to model outputs. CNBC reported July 1 that China-linked actors have targeted AI startups through cyberattacks, insider threats and espionage as competitive pressure in the sector intensifies. CNBC

What emerges is a genuinely two-sided restriction effort taking shape simultaneously in Washington and Beijing, each government moving to limit the other's models and chips from crossing its borders, even as commercial incentives on both sides keep pulling in the opposite direction. Cost pressure and open-weight distribution are powerful enough forces that policy alone has struggled to contain them so far — Lindy's cost-driven switch to DeepSeek and Coinbase's disclosed use of two Chinese models illustrate that enterprise procurement decisions are, for now, running ahead of regulatory intent on both sides of the Pacific.