AIToday
Large Language ModelsAI Regulation & PolicyArs Technica AIPublished: Sep 19, 2026, 06:00 JST

Federal Register site drops Alibaba Qwen search tool

Federal Register site drops Alibaba Qwen search tool

3 Key Points

  1. What happened

    On Wednesday, US government officials removed a Chinese Alibaba Qwen AI search tool from the Federal Register website, after a widely shared screenshot from user "tleilax___" showed the option on September 15.

  2. Why it matters

    The removal came after social media users noticed an apparent contradiction, since top US law enforcement had earlier alleged such models illegally copy US frontier models. The service was available for at least a day.

  3. What to watch

    Whether the government use of the Qwen model posed a security risk depends on how it was trained, Georgetown's Anupam Chander told Reuters, and whether any US data was processed by Alibaba-controlled systems, per Senator Mark Warner. Experts told Reuters it likely posed no substantial risks.

WHO IT HITSUS government agencies and their IT teams that download open-weight AI models for local search now face conflicting signals: experts say such use can give more data control, while some lawmakers argue no federal entity should use any Chinese AI model.

Not sure about something? Ask the AI

Summaries like this, in your inbox every morning.

Context & Analysis

The episode surfaced a tension that had been building for weeks. Earlier this month, the FBI named Alibaba among six leading Chinese firms allegedly conducting "industrial-scale distillation"—a practice the agency says helps America's biggest competitor cut costs and development time in the global AI race. Against that backdrop, the Federal Register's use of a Qwen model, however small, struck some as a direct contradiction. Daniel Castro, president of the Information Technology and Innovation Foundation, told Reuters the disconnect was "insane" given the FBI's encouragement to use only American models.

Yet the technical details suggest the risk was limited. The model in question was a small open-weight version of Qwen3 0.6B level, not Alibaba's flagship, and it did not send government data to Alibaba or third parties. Security experts told Reuters that for US agencies, such downloadable models can actually give more control over data because they run locally rather than sending queries to external systems. It is unclear when the National Archives, which runs the Federal Register, began offering the option, and the agency has not commented on the removal.

What happens next may hinge on how Washington resolves a broader question: whether AI models that originate in China but are controlled by American institutions should face the same restrictions as other Chinese services considered higher security risks. Representative John Moolenaar (R-Mich.), who chairs the House China Committee, has said no federal entity should use a Chinese AI model. Meanwhile, business owners who rely on open source models have urged the Trump administration not to restrict access to Chinese models, and policy research submitted to the US-China Economic and Security Review Commission in March warned that US export controls may be targeting the wrong layer of the competition. For federal IT teams, the practical test is whether local, open-weight models remain a viable option or become politically untenable.

FAQ
Why was the Qwen AI search tool removed from the Federal Register?
Social media users noticed the site was using an Alibaba Qwen model even as US law enforcement claimed such models illegally copy US frontier models. US officials then removed it on Wednesday.
Did the Qwen model on the Federal Register pose a security risk?
Experts told Reuters it was unlikely to have posed substantial risks. Georgetown's Anupam Chander noted the site's content is already public, so the model was not working with sensitive government information.
What kind of Qwen model did the Federal Register use?
The site used a small open-weight model of "Qwen3 0.6B level," not Alibaba's largest flagship model. It was open source, served solely to retrieve documents, and did not send government data to Alibaba or third parties.
Ars Technica AIRead Original Article

Get the latest Large Language Models news every morning

For example, today's edition would include:

  • Goldman Sachs: consumer agents mark new AI growth phaseTop Companies AI · 1h ago
  • Mastercard rolls out AI-agent payments; only 7% would let a bot buyTop Companies AI · 1h ago
  • AWS's Matt Wood: AI is repeating the cloud playbookTop Companies AI · 1h ago

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. Q&As are published on this page for other readers too.

Related Articles

Next articleAmazon SageMaker AI adds 13 inference capabilities in 2026