Federal Register Used Alibaba's Qwen for Federal Rulemaking Search
Just one week after the FBI accused Alibaba of copying Anthropic's models, Qwen-powered search was discovered on the Federal Register — and removed the same day.
Federal Register Used Alibaba's Qwen for Federal Rulemaking Search
Just one week after the FBI accused Alibaba of copying Anthropic's models, Qwen-powered search was discovered on the Federal Register — and removed the same day.
Federal Register, the official publication for US federal rulemaking, used a search function powered by Qwen, the AI model from China's Alibaba, to search proposed federal rules. That is according to Reuters reporter Courtney Rozen's report from September 17. The search function was removed on Wednesday, September 17, at roughly the same time that posts about it appeared on social media, according to screenshots and an archived version of the page's source code reviewed by Reuters (source).
A Timed Contradiction
The chronology gives the story its force. The week before, the FBI claimed that Alibaba had copied the models of its American rival Anthropic. US officials said at the time that Chinese AI companies were engaged in "malicious" copying of American technology at "industrial scale" — an accusation from US authorities that is neither confirmed by Alibaba nor commented on by Anthropic in the source material.
Barely a week later, Qwen was found running on a US government website, and was removed the same day it became known. The removal also came in the days before a planned leaders' meeting between the US and China. There is no documentation in the source of any connection between the removal and the meeting, but the contradiction between the official accusation and the actual use of Qwen in federal administration became immediately visible.
How the Finding Was Made
It was not the government itself that uncovered the usage. Social media posts drew attention to the search function, and Reuters documented it using screenshots and an archived version of the Federal Register's source code. When the tool was deployed is unknown — there is no confirmation of how long the Qwen-powered search was in use, who approved it, or whether it was part of a formal procurement arrangement.
How Serious Was It? The Experts' Assessment
Both expert commentary and political reactions point in the same direction: the actual risk depends entirely on technically unknown details.
Anupam Chander, a professor at Georgetown Law, believes the use of Qwen does not appear to pose an immediate cybersecurity risk. The argument is twofold: the Federal Register's content is already public, so the model was not working with sensitive government information. Moreover, open models like Qwen can run on an organization's own infrastructure. But Chander reserves judgment on this depending on how the model was trained — a fundamental question that remains unanswered.
Senator Mark Warner, a Democrat and vice chair of the Senate Intelligence Committee, said in a statement that the risk depends on whether American data left the government's security perimeter and was processed on Alibaba-controlled systems. That question remains unanswered: it is neither confirmed that the data was sent to Alibaba, nor that the model ran internally within the government.
Not the First Time Qwen Has Been on the Agenda
The Federal Register is not the first case in which Qwen's use in American services has drawn attention on Capitol Hill. In April, the House committees on China and homeland security sent a letter to Airbnb, which also uses Qwen, requesting details about its use — as part of an investigation into "the national security risks created by the integration" of Chinese AI tools into platforms that Americans use.
The April investigation shows that political attention to Chinese AI tools in American services has a broader focus than Alibaba alone, extending beyond the federal administration.
The Unanswered Questions
None of the parties involved has provided explanations. The National Archives, which operates the Federal Register, and the White House have not responded to inquiries. The FBI declined to comment on Thursday. Alibaba is not heard from in the source material.
A number of fundamental questions remain: When was the Qwen search deployed, and how long was it in operation? Who approved its use? Were search queries processed on the government's own infrastructure, or were they sent to Alibaba-controlled systems? Were any American data exposed to security risk?
Until these questions are answered, the case remains primarily an example of the gap between Washington's China rhetoric and the practical reality within its own administration.