LLM are becomming widely used and whoever controll the LLM training control people lives to some extends. It's crazy how people are relying more and more on ChatGPT for conveniance, some even think it is actually intelligent/smart (it is not). LLM have biais because training data and training recipes are not neutral. Governments and big corpos can use this to extend control and manipulate/influence opinion, behavior, etc... They can choose how and what to censor, the point of view of information that could change user perception of it. With closed data and training recipe we can not see how it is made and what rules the models have been following.

Most open source model only publish the finally wieght model. No training data (mainly because they violate IP, copyright, etc... if you ask me) nor the training recipe, so you only have the possibility to host the model yourself. They slap Open Source on it and voilà... but it often isn't and when it is you can see that it perform worse (because they haven't read all Sci-Hub and Z-Library unlike Llama and ChatGPT for exemple.

Edit : Clarify some sentences

Your tl;dr appears to be missing some important data. You can have an opinion but please don't represent it as an accurate summary.

Things you crucially missed:

• Less open than every other service available
• Bills itself as the most open
• Server side source code is MIA
• No model card available. Evaluations, risks, biases, guardrails and safety measures unclear.

In the 1 September update they state they found that web client and a mobile client as well, but not the API (I guess) containing the system prompt and the actual routing to the models.

It supports the thesis that Lumo is not open source in many common sense ways that most people would expect when a model claims it is open source. So in that sense, it does though.

Can you be more specific?

How so? It's pretty clear that Proton using "open source" in context of their LLM service is demonstrated to be false.