It's not a bug, it's a feature Large language models, or LLMs, are biased in one way or another - often many. And there may be no way around that....