Glonce

Alibaba's Qwen removed from US government site after backlash

policyThe Glonce editorsArs Technica🔥65Glonce rating65Impact60Novelty45Relevance85Surprise60Players70Credibility82
Alibaba's Qwen removed from US government site after backlash

On Wednesday, US government officials removed a Chinese AI search tool from the Federal Register website, Reuters reported. The Federal Register site is run by the National Archives, an independent agency tasked with public access to federal documents, and the tool let visitors search public comments on proposed regulations using one of Alibaba's Qwen models. It is unclear exactly when the National Archives began offering it.

The removal followed a wave of attention on social media, where users pointed out an apparent contradiction: a US government agency was running an Alibaba AI model at the same time that top US law enforcement was warning against Chinese models. Earlier this month, the FBI had named Alibaba among six leading Chinese firms it says are conducting 'industrial-scale distillation,' meaning they allegedly copy US frontier models illegally in ways that help China cut costs and development time in the race for AI leadership. A user with the handle 'tleilax___' posted a widely shared screenshot on September 15 showing the Qwen option live on the government site. Reuters said an archived copy of the site's source code confirmed the model was taken down on Wednesday, meaning the tool had been available for at least a day.

Reaction focused on the mismatch between policy and practice. Daniel Castro, president of the Information Technology and Innovation Foundation, called it an 'insane' disconnect for a US agency to run an Alibaba model while the FBI tells stakeholders to stick to American ones. Georgetown University law professor Anupam Chander told Reuters that whether the government's use of the Qwen model actually posed a security risk depends on how the model was trained. Senator Mark Warner (D-Va.) added that it also matters whether any US data was processed by 'Alibaba-controlled systems.'

The National Archives has not commented on the removal and did not respond to Ars Technica's request for comment; neither the White House nor the FBI has commented either. It remains unexplained why the agency adopted the Qwen-based tool in the first place.

Key facts

  • The National Archives ran an Alibaba Qwen AI search tool on the Federal Register website, letting visitors search public comments on proposed regulations.
  • A screenshot posted on X by user 'tleilax___' on September 15 showed the option live; officials removed it the following Wednesday, and archived site source code shows it was available for at least a day.
  • Earlier this month the FBI had named Alibaba among six leading Chinese firms it accuses of 'industrial-scale distillation,' alleging they illegally copy US frontier models to help China's AI race.
  • ITIF president Daniel Castro called the mismatch 'insane'; Georgetown professor Anupam Chander said the security risk depends on how the model was trained; Senator Mark Warner flagged whether US data reached Alibaba-controlled systems.
  • The National Archives, the White House, and the FBI have all declined to comment on the removal or on how the tool came to be deployed.

Why it matters

A federal agency ran a model from a company the FBI had just accused of illegally copying US AI technology, while the same FBI was telling other agencies to use only American models. The episode shows a gap between the government's declared AI-security posture and what one of its own websites was actually running, and it surfaced only because outside social media users noticed, not through any internal review.

Who it affects

The National Archives and the Federal Register site it runs; visitors who used the search tool to look up public comments on proposed regulations; the FBI's credibility on its distillation warnings about Alibaba and other named Chinese firms; and Alibaba's Qwen models, whose US government use is now public.

How to use it

There is no product or price here for readers to act on; the practical takeaway is for agencies and vendors. Before deploying any AI feature on a public-facing government site, the vendor's country of origin should be checked against active security guidance, since this tool apparently went live without that check happening first.

How solid is it

The account comes from Reuters via Ars Technica and rests on a verifiable trail: a public screenshot dated September 15 and an archived copy of the site's source code showing the removal on Wednesday. Named sources (Castro, Chander, Warner) are quoted directly. What is missing is any confirmation from the parties actually responsible: the National Archives, the White House, and the FBI all declined to explain when or why the tool was added.

Risks and caveats

The story does not establish that any US data was actually processed by Alibaba-controlled systems, only that Senator Warner said it would matter if it had been. Nor does it confirm a security incident occurred; Chander's comment leaves that contingent on how the model was trained, which is not stated. The tool's exposure window is bounded only as 'at least a day,' and no exact date for its introduction is given.

“insane”

— Daniel Castro, president of the Information Technology and Innovation Foundation