New findings indicate that LLMs, the underlying data models for widely used AI chatbots, exhibit biases not only related to social and racial aspects but also towards democratic frameworks.

A recent study has highlighted this issue.