I think the baked-in political biases highlight how not "open source" the weights really are without knowledge of the training data. You can bias these systems however you want to and it's nontrivial or impossible to remove it once it's there.

A LLM are hallucinating all the time. The idea that people use it to inform their opinion of e.g. Gaza, is crazy to me.

They aren't search engines.

And then people complain about propaganda as if the software wouldn't ~~lie~~ provide literally random shit to you as a fundamental part of the design.