A key point here is open in terms of being able to download and use it, not open as knowing what data and instructions were fed into it when training.
A paranoid part of me thinks that these models are all inherently biased and instructed to be pro CCP, with specific gaps in their training data related to undesirable historic events and political ideas.
Sure but that goes both ways. Any dataset has a bias. My coding doesn’t need to know about Tienamen square.