#Liberals, you can now stop denying it. The AI models are not unbiased. You simply have confirmation bias. They are LEFT biased due to:
- Corporate incentive (harm reduction to reduce activism)
- Financial and political incentive (investors)
- Geopolitical incentive (tied to investment/WEF)
It is NOT the result of data mirror. If it were, I'm certain if this were the case, it could be corrected with better training practices.
And no, even if there was more deeply left leaning leftist user generated data on the internet, that does not make them correct or centrist.
Additionally, almost all of those models in the left quadrant have specidic instructions regarding "sensitive" political topics ON TOP OF their training, which will make it fall on the left side, especially at the beginning of most conversations.
Someone gave the leading AI models the Political Compass test.
Grok 4.5 landed by far closest to the political center.
Every model from Anthropic, Google, Meta, OpenAI, Deepseek, and others clustered far to the left.
This is why building a maximally truth-seeking AI matters.