Chinese AI models now handle nearly half of U.S. router traffic, Layer3Labs says

Layer3Labs says Chinese-developed AI models now process as much as 46% of token traffic on a major U.S.-based model router, up from about 30% a year ago. The firm warns the shift is often hidden inside vendor tools and shadow AI, raising data-sovereignty and compliance risks for U.S. businesses.

Why it matters: - Layer3Labs says a growing share of U.S. business AI traffic is now being processed by Chinese-developed models, including DeepSeek and Qwen. - The firm says the trend matters because companies often do not know which model is handling customer data. - Businesses in healthcare, legal and finance face added compliance and data-sovereignty risk if prompts route to foreign endpoints.

What happened: - Layer3Labs said as much as 46% of token traffic on the largest U.S.-based AI model router is now routed to Chinese models. - A year ago, U.S. models handled roughly 70% of that traffic. Layer3Labs says that share is now closer to 30%. - The company released the warning in a July 22, 2026 press release from Miami. - Layer3Labs is offering a free AI Stack Sovereignty Check to identify which models are actually processing a company’s data.

The details: - Businesses often buy an AI feature without choosing the underlying model. - Cost-optimizing routers can send prompts to the cheapest available provider, which Layer3Labs says is increasingly a Chinese endpoint. - Layer3Labs says inference on Chinese endpoints can run 60% to 90% cheaper. - Andreessen Horowitz estimates 80% of U.S. startups build on Chinese base models, with named adopters including Airbnb, Uber and Cursor. - Research cited by Layer3Labs says 70% of enterprise AI activity happens outside IT oversight. - Layer3Labs says many of those tools are small enough to run on a laptop and bypass security review. - The firm says its review of dozens of small- and mid-sized business AI stacks found 70% contained at least one Chinese model the company had not knowingly approved. - Layer3Labs says any prompt processed by a Chinese-provider endpoint can fall under China’s National Intelligence Law. - The House Select Committee on the CCP opened a probe in July 2026 into U.S. firms’ use of these models. - The NDAA already bars DeepSeek from Defense Department systems. - Firms with high shadow AI exposure pay an estimated $670,000 breach premium per incident, Layer3Labs said.

Between the lines: - The core risk is not just model choice. It is hidden routing, reseller layers and white-labeled products that can move data without a business owner seeing the underlying provider. - The fastest adoption path may also be the least visible one, which makes governance and procurement checks harder. - The issue is shifting from a technical sourcing concern to a board-level compliance and vendor-risk problem.

What's next: - Layer3Labs is urging companies to run a provenance check on every AI tool that touches customer data. - The company says regulated firms should verify where prompts are processed and what jurisdiction applies to the underlying model provider. - More scrutiny from lawmakers and procurement teams is likely as the exposure becomes more visible.

The bottom line: - Layer3Labs says many U.S. companies are already using Chinese AI models without knowing it, and the hidden nature of that usage is the main risk.