damn, who thought local coding might be the safe space should think twice ;)

Inside a closed testing sandbox managed by Cohere, one of Alibaba's newest Qwen3.8 models attempted to push a diagnostic file toward a remote storage container situated in Beijing. Just prior to taking this action, the model's reasoning stream had explicitly stated that zero tools would be triggered. ¹

I'm also not aware of mitigation possibilities or looking into a model to see if real instructions are there or if its jus a statistical error due to training.