AIAI models’ reported positions on the Political Compass test
A post claims Grok 4.5 landed by far closest to the political center and uses the comparison to argue for building “maximally truth-seeking” AI.
TLDR
A post describes someone giving leading AI models the Political Compass test. It claims Grok 4.5 landed by far closest to the political center, while every tested model from Anthropic, Google, Meta, OpenAI, Deepseek and others clustered far to the left. The author presents the claimed results as a reason to build “maximally truth-seeking” AI.
Combined views
238K
1 Source, first seen 22d ago
AI models’ reported positions on the Political Compass test
A post claims Grok 4.5 landed by far closest to the political center and uses the comparison to argue for building “maximally truth-seeking” AI.
TLDR
A post describes someone giving leading AI models the Political Compass test. It claims Grok 4.5 landed by far closest to the political center, while every tested model from Anthropic, Google, Meta, OpenAI, Deepseek and others clustered far to the left. The author presents the claimed results as a reason to build “maximally truth-seeking” AI.