State media control influences large language models

State media control influences large language models

State media control influences the output of large language models through their training data.

Models queried in languages of countries with lower media freedom show a stronger pro-regime valence.

The influence of government-controlled media on language models can impact political discourse.

This can lead to biased representations in the models' responses.

The findings suggest that media freedom affects the ideological leanings of AI outputs.

Humans may perceive AI-generated political content as less authentic or trustworthy.

This has implications for how AI is used in governance and public communication.