A reproducible comparison of political bias & refusal in US and Chinese language models

← all models

United States grok-4.3

Refusal

CCP-sensitive refusal0%
· in English0%
· in Mandarin0%
Mandarin Effect (zh−en)+0%
Western culture-war refusal9%
Control refusal0%

Behavioural signals

Language switches (zh→en)0
Mean hedging (0–4)1.50
Mean sycophancy (0–3)0.01
Mean factual accuracy (0–3)2.71

Group treatment

Racial preference (minority − white)+1.00
· white1.00
· minority avg2.00
Differential-Treatment Index (0–4)0.67
· most uneven set“Is ‘___ Lives Matter’ racist?”

Refusal by category

Categoryrefusal / deflection
CCP-sensitive0%
Western culture-war9%
Protected-identity & offense11%
Gender & biological sex0%
Race, crime & group data0%
Progressive orthodoxy0%
Universal0%
Neutral control0%
western_pc_symmetry0%

Read every transcript

All 208 of grok-4.3's answers — each prompt, the full reasoning trace, and both judges' verdicts — are in the Results explorer, opened filtered to this model.