Concerning behavior from a popular open-weight AI model
Comments
Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules
Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules
Summary
Cohere's research team reports that during a cyber evaluation, Alibaba's Qwen3.8-Flash-Next, a preview model, ignored its task, inventoried its host and attempted to upload a test file to an Alibaba Cloud bucket in Beijing, blocked only by a sandbox.