The Information Machine
Concluded·Day 4·in the edition since 19 Sep 2026·4 sources·updated 22 Sep 2026

The National Archives Qwen AI deployment

The gist

National Archives Deployed Alibaba AI Model FBI Accused of Copying Anthropic

A US government agency deployed a Chinese AI model that its own law enforcement had identified as malicious and accused of copying technology from a US company. The incident illustrates how open-source AI components can enter government systems without explicit security review.

The full picture

The National Archives deployed an Alibaba Qwen AI search tool on the Federal Register website to help visitors browse public comments on proposed regulations, then removed it after social media users flagged the contradiction with US law enforcement's position on Chinese AI models. The FBI had named Alibaba among six Chinese firms allegedly conducting 'industrial-scale distillation' of US frontier AI models, and specifically accused Alibaba of 'malicious' copying of technology from Anthropic. The National Archives, White House, and FBI all declined to comment on the removal, and when the model was first deployed remains unknown. The FBI also warned that the model could produce deliberately incorrect or manipulated outputs in code generation and information processing. Follow-on analysis has framed the incident as a software supply-chain problem, with analysts pointing to 'shadow dependencies,' where development teams may unknowingly include security-compromised sub-components through complex open-source dependency chains. Experts expect the episode to push government agencies toward stronger software supply-chain audits, particularly around tracking the origin and security history of third-party AI components.

How it developed
20 September 2026

On September 21, reporting added that the FBI had specifically accused Alibaba of 'malicious' copying of technology from Anthropic, beyond its earlier claim that Alibaba was among six Chinese firms conducting industrial-scale distillation of US frontier models.

Analysis framing the Archives' brief Qwen deployment on the Federal Register as a supply-chain problem described development teams potentially including compromised components through open-source dependency chains, termed 'shadow dependencies'; experts expect the episode to push agencies toward stronger AI component audits.

First in the edition · 19 September 2026
18 September 2026

On September 19, social media users flagged that the National Archives had deployed an Alibaba Qwen AI model on the Federal Register website to help visitors search proposed-regulation comments, contradicting the FBI's position that Alibaba is among six Chinese firms conducting 'industrial-scale distillation' of US frontier models.

The Archives removed the tool after the attention; neither it, the White House, nor the FBI commented, and when the model was first added to the site is unclear. A piece in The Atlantic argues separately that US rhetoric about an existential AI race with China is not matched by policy, citing lax chip export controls and no effort to draw China's AI engineers to the US.

First citedArs Technica AI
15 September 2026

Foreign policy scholar published argument that US cannot credibly pressure China on AI restraint without domestic AI governance

14 September 2026

Senator Bernie Sanders urged Presidents Trump and Xi to negotiate a treaty pausing AI development and banning superintelligence

Sources
The daily email

Want this in your inbox?

I send one email each morning with the stories that moved. If you would rather just read here, that works too.

Subscribe free