I’ve seen a lot of talk about how devices get ‘compromised’ running Deepseek. Two ways I’ve heard of so far. One, download the app without reading the privacy policy and willingly give your stuff to China. Second, accidentally download a malicious Python package associated with Deepseek.
Thusfar, unless someone can link a cyber report here, I have yet to see anyone show that running the open source model of Deepseek compromises your computer. Whether that is via Ollama, LM Studio, etc, no evidence showing that the model reaches outside the service it is running on.