Language models may reflect the biases and stereotypes present in their training data.

Language models are designed to generate human-like language.

Language models are not capable of generating biased text.

Bias in language models only affects certain applications.