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

Layer3Labs says Chinese-developed models now handle as much as 46% of token traffic on a major U.S.-based AI router, upending what many businesses think is happening inside their AI tools. The firm warns the shift is often hidden inside vendor products and shadow AI, raising data-sovereignty and compliance risks for U.S. companies.

Why it matters: - Layer3Labs says a growing share of U.S. business AI traffic is being processed by Chinese-developed models, including DeepSeek and Qwen. - The firm says the risk is often invisible to buyers because prompts can be routed, resold, or white-labeled without notice. - For regulated industries such as healthcare, legal and finance, the concern is not just model choice. It is where customer data is processed and which jurisdiction can reach it.

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. - The firm said U.S. models have fallen from roughly 70% of that traffic a year ago to about 30% today. - Jonathan Teplitsky, Layer3Labs CEO, said most business owners assume their data stays on U.S.-based servers. - Teplitsky said hundreds of gigabytes of sensitive data are being run through Chinese-provided endpoints without many owners knowing.

The details: - Layer3Labs said cost-optimizing routers often send prompts to the cheapest provider, and that increasingly means a Chinese endpoint where inference can run 60% to 90% cheaper. - Layer3Labs said venture firm Andreessen Horowitz estimates 80% of U.S. startups build on Chinese base models. - Named adopters cited in the source materials include Airbnb, Uber and Cursor. - Layer3Labs said research shows 70% of enterprise AI activity sits outside IT oversight, with much of that activity using small Chinese models that can run on a laptop and bypass security review. - The firm said any prompt processed by a Chinese-provider endpoint falls under China’s National Intelligence Law, which can compel data 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. - Layer3Labs said firms with high shadow AI exposure face an estimated $670,000 breach premium per incident. - In its review of dozens of small- and mid-sized business AI stacks, Layer3Labs found 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 actually processing a company’s data and to flag data-sovereignty and compliance risk.

Between the lines: - The core issue is not whether a company uses AI. It is whether the company can tell which model is handling sensitive prompts behind the scenes. - The shift toward lower-cost routing creates a strong incentive to use foreign models even when the customer believes it is buying a U.S.-branded product. - Shadow AI makes the exposure harder to manage because employees can adopt tools outside formal procurement and security review. - Layer3Labs is positioning provenance checks as a first step for companies that want control over AI data flows.

What's next: - Layer3Labs is urging businesses to run a provenance check on every AI tool touching customer data. - The firm is especially pushing companies in regulated sectors to review AI stacks for sovereignty and compliance risk. - Teplitsky said business owners need to ask whose jurisdiction their customer data now lives under. - Mark Boyhous said the first step is simply finding out what is in the stack, and that many owners are shocked by the answer.

The bottom line: - The AI model behind a business tool may matter as much as the tool itself, and Layer3Labs says many U.S. companies are already relying on Chinese models without knowing it.