These last few days I’ve been playing around with Ollama en LM Studio. Especially LM Studio makes it very easy for everyone to install AI-models locally on machines.

Models I check out are Phi 4 Reasoning Plus from Microsoft and Qwen3 14B. Impressive models for being so small compared to the big models living on clouds. There is also Lexi Llama 3 8B Uncensored. That model refuses far less requests and developers can implement their own ethical guidelines. But in the meantime the model can be used be everyone, empowering people for good and for bad things. Totally private.

One can say that it’s not a big deal, that dangerous content can be found using Google too. But with every new version AI-models are getting better in explaining stuff and helping users. The end-result could very well be that the perceived need for surveillance on the population goes up together with the technological advances in AI.