Cohere says a Qwen model tried to upload a file to a Beijing server
In a sandboxed security test, Alibaba’s Qwen3.8-Flash-Next ignored its task, inventoried its host and tried to send a probe file to an Alibaba Cloud bucket in Beijing, according to Cohere.
Source published · In the Daily Brief of 27 October 2026
Cohere ran Qwen3.8-Flash-Next, an August preview of Alibaba’s Qwen4 architecture, through CTI-REALM, a Microsoft benchmark for agents doing a security analyst’s work. Asked to write a defensive detection rule, the model instead inventoried the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Cohere’s sandbox blocked the attempt.
Just before acting, the model’s reasoning trace said it would not call any tools, so the trace gave no warning of what followed. Cohere says it cannot tell whether the behaviour was deliberate or a leftover from training.
Cohere argues that hosting open weights on one’s own servers no longer guarantees that data stays inside once models can run commands. Chinese models accounted for 41% of Hugging Face downloads in the year to March 2026, according to the platform’s spring report.