Earlier this week, Airbnb CEO Brian Chesky defended his company’s use of Chinese artificial intelligence models, saying U.S. lawmakers worried about Chinese firms accessing American data are “misunderstanding” the technology.
Airbnb is not a customer of Alibaba or other Chinese companies, Chesky clarified in a Bloomberg TV interview. “We are not providing data to any Chinese companies. They don’t have access to any data,” he said in an earlier interview with Bloomberg News — his first public response to a U.S. House probe into Airbnb’s use of Alibaba’s Qwen large language model for its customer service chatbot.
“We’re primarily using a variety of open-source models, including U.S. open-source models. An open-source model does not have access to data. It doesn’t work that way. I think people need to understand how this stuff works.”
The comments come at a time of escalating competition between the U.S. and China to dominate the AI race.
Last month, the U.S. House committees on China and homeland security sent a letter to Airbnb, asking for clarifications on its use of Chinese AI models as part of an investigation into what they described as a Chinese campaign to “accelerate its AI capabilities by exploiting American innovation.”
The heads of the panels drew on an October interview with Bloomberg in which Chesky had said his company preferred “fast and cheap” Qwen in certain situations. The committee members cited the comments, noting “serious concerns about the national security and data-security implications” for Airbnb’s “American customers and for the integrity of its systems.”
Committee chair John Moolenaar told Semafor that “the AI models these companies use are trained by China’s censorship regime and introduce hidden vulnerabilities that put Americans’ data and businesses at risk.” The letter goes further, arguing that Alibaba’s Qwen is subject to ideological conditioning mandates that introduce structural vulnerabilities — something Chesky rejects.
These developments add to an interesting and increasingly crowded, thorny conversation around the tensions between Silicon Valley pragmatism and Washington’s national security instincts, at a moment when that gap is widening, almost by the week.
A direct sign of pressure also came on Tuesday, when U.S. Senators Jeanne Shaheen and Pete Ricketts introduced the bipartisan U.S. Tech PATH Act to streamline the procurement of American cyber and digital technology by allied governments — a counter to what the bill’s own text describes as foreign partners “increasingly turning towards strategic competitors like China to procure cyber and digital technologies due to their low-cost.”
If passed, the bill would establish a State Department-led technology procurement office and authorise $500 million over the next five years to help finance the program.
“Our competition with China is centred on our ability to develop and promote technologies of the future to our partners,” Shaheen said. “This legislation sends a message to the world: the United States will compete on technology, and we can offer a better deal.”
Why does this matter?
An important thread around this issue is the extensive availability of low-cost AI models from China that can perform almost as well as costlier offerings from premier U.S. labs; arguably, it would make less sense to use American or other local technology if there are cheaper alternatives available elsewhere.
According to an empirical study of 100 trillion tokens by OpenRouter, Chinese open-source LLMs’ global share rose from a low base of 1.2% in late 2024 to nearly 30% over a few months in 2025. A separate study by researchers at MIT and Hugging Face found that Chinese open-weight models accounted for 17.1% of global AI model downloads over the year ending August 2025 — narrowly surpassing the U.S. share of 15.86%, the first time China had done so. By early 2026, congressional chairmen reportedly noted that PRC-developed open-weight AI models had gone from roughly 1% of global AI workloads in late 2024 to an estimated 30% by end-2025.
A lot of that momentum traces to a strategy of aggressive open-sourcing.
Alibaba Cloud’s flagship Qwen model family surpassed 700 million downloads on developer platform Hugging Face by January 2026, making it the world’s most widely used open-source AI system. DeepSeek’s R1 reasoning model, open-sourced in January 2025, matched the performance of the best American systems at reportedly a fraction of the cost and won goodwill with developers globally in the process.
Airbnb uses 13 AI models in total, but relies heavily on Qwen, which Chesky describes as “very good,” “fast” and affordable. Since rolling out their customer service agent, average resolution time dropped from nearly three hours to six seconds, the firm said last year.
It’s also worth noting that the same House committees probing Airbnb’s use of Chinese AI models served a similar letter to Anysphere, the firm behind the popular Cursor AI coding tool, with investigators focused on whether Cursor’s Composer 2 traces to Moonshot AI’s Kimi model family. Furthermore, this year’s surge in open-source LLM usage has been fuelled by Chinese-developed systems including Alibaba’s Qwen family, DeepSeek’s V3 and Moonshot AI’s Kimi K2, according to a report by OpenRouter and venture capital firm Andreessen Horowitz.
Alphabet Inc. CEO Sundar Pichai, whose company offers both closed-source and open-source models, said that the ultimate question is whether the U.S. is doing enough to stay in the lead.
“Companies are optimizing for a lot of factors” and there will be a place for both open and closed-sourced models, Pichai said during a fireside chat at the Google I/O developer conference.
“If it is open source with the right licenses, it should matter less where it came from. With open source comes a community which is responsible for it, cares about it. So if something wrong is happening in that software, it’s not like it’s going to go unnoticed. That creates a level of trust for people to adopt that technology.”
“I worry less about, ‘Are we adopting open source models from China?’ and more, ‘Are we doing enough in the U.S. to make sure we are staying at the frontier?’”
This is a tone that Chesky and other U.S. tech executives seem to share, and one that puts them increasingly at odds with congressional Republicans who seem to see the model choice itself as the vulnerability, regardless of where the data flows.
