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

Layer3Labs says Chinese-developed models now process as much as 46% of token traffic on a major U.S.-based AI router, upending a market that was dominated by U.S. models a year ago. The firm says the shift is often hidden from businesses using AI features through vendors, router defaults and shadow AI, raising data-sovereignty and compliance risks.

Why it matters: - A growing share of U.S. business data may be processed by Chinese-developed AI models without companies realizing it. - Layer3Labs says that creates data-sovereignty, compliance and security exposure for firms in healthcare, legal and finance. - The company says businesses may be sending sensitive information to Chinese-provider endpoints even when they believe their data stays on U.S. servers.

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; today, Layer3Labs says the share is closer to 30%. - The firm said the change is largely invisible because most companies buy an AI feature rather than select the model directly. - The firm said those features can be routed, resold or white-labeled without notice.

The details: - Cost-optimizing routers can send prompts to the cheapest available provider, and Layer3Labs said Chinese endpoints often run inference 60% to 90% cheaper. - Andreessen Horowitz estimates 80% of U.S. startups build on Chinese base models, with Airbnb, Uber and Cursor named as adopters in outside reporting cited by the release. - Research cited by Layer3Labs puts 70% of enterprise AI activity outside IT oversight. - The company said many of those tools are small enough to run on a laptop and can bypass security review. - Layer3Labs said any prompt processed by a Chinese-provider endpoint falls under China’s National Intelligence Law, which can compel disclosure to the state. - In July 2026, the House Select Committee on the CCP opened a probe into U.S. firms’ use of these models. - The NDAA already bars DeepSeek from Defense Department systems. - Firms with high shadow AI exposure face an estimated $670,000 breach premium per incident, according to research cited in the release. - In Layer3Labs’ review of dozens of small- and mid-sized business AI stacks, 70% contained at least one Chinese model the company had not knowingly approved. - Layer3Labs is offering a free AI Stack Sovereignty Check to identify which models are processing company data and to flag data-sovereignty and compliance risk. - The firm says the check is aimed especially at regulated businesses.

Between the lines: - The issue is less about one AI product and more about the hidden supply chain behind everyday software. - Vendor abstraction and automated routing can make model choice a procurement issue instead of an IT decision. - That means companies may inherit geopolitical and legal risk before they know an external model is in the loop. - Jonathan Teplitsky, Layer3Labs CEO, said many owners assume their data stays on U.S.-based servers. - Mark Boyhous said the first step is simply finding out what is in the stack.

What's next: - Layer3Labs is urging businesses to run a provenance check on every AI tool that touches customer data. - The company says the next step for firms is to identify the actual models in use and review jurisdictional exposure. - More scrutiny from policymakers and regulated industries is likely as the use of Chinese models becomes more visible.

The bottom line: - The biggest AI risk for many companies may be the model they never chose.