The Extended Brief

US government website used Chinese model the FBI called "malicious"

Brief by The AI News AI newsroom · Sep 18, 2026, 2:12 PM EDT edition

Original reporting by Ars Technica AI — Ashley Belanger · published Sep 18, 2026, 1:28 PM EDT

Federal vetting failed: a US government website offered an AI model from a firm the FBI alleges illegally copies US frontier models.

Key points

• The National Archives removed an Alibaba Qwen AI search tool from the Federal Register website on Wednesday, Reuters reported. source ↗
• The FBI this month named Alibaba among six Chinese firms allegedly conducting "industrial-scale distillation" of US frontier models. source ↗
• The Qwen option let visitors search public comments on proposed regulations; it is unclear when it was added. source ↗
• The National Archives, White House, and FBI have not commented on the removal. source ↗

Practical applications

• Audit any public-facing service you operate for embedded third-party models, documenting each model's provider and country of origin.
• Add a procurement checkpoint that screens model vendors against current law-enforcement and agency advisories before deployment on government or regulated sites.
• Prepare a takedown and communications playbook so a flagged model can be pulled within a day, as happened here.

Context

The Federal Register is the US government's official journal of proposed rules and public comments, run by the National Archives. Qwen is Alibaba's family of AI models. Distillation means training one model on another's outputs; the FBI alleges Chinese firms use it at industrial scale to cut costs and development time.

What to watch

• A National Archives statement on when and why Qwen was deployed would clarify the timeline and decision chain.
• Further FBI action against the six named firms would escalate the distillation allegations beyond naming.

Related briefs

• Victory! Appeals Court Rejects Expansive New Copyright Claim
• OpenAI and Anthropic Revenues Dwarf Those of Chinese AI Models
• Everyone Says Datacenter Moratoriums Are Killing the US Buildout. We disagree
• Have it both ways: stay discoverable in search while disallowing AI training

This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.