Vals study finds Chinese AI models' pro-government lean varies by model and prompt language
Overview
Vals.ai tested Chinese and non-Chinese AI models and found their pro-government lean was lower than expected overall, according to Vals, though the lean differed widely between individual models.
For most models tested, answers in Simplified Chinese were more pro-government than answers in English, Russian, or Spanish, and Vals reports the effect was much stronger for Chinese models.
Refusals were scored as neutral. Many Chinese models were tested through third-party providers rather than their official services.
Written by AI from the articles below · updated Nov 1, 7:37 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- The DecoderNewsChinese AI models show politically complex behavior, not simple government mouthpieces
AIVals.ai tested Chinese and non-Chinese models and found their pro-government lean was lower than expected, though it varied widely by model. The slant rose when prompts were in Simplified Chinese, and Vals reports that the effect was much stronger for Chinese models. Refusals counted as neutral in the scoring, and many Chinese models were tested through third-party providers rather than their official services.
Heat Trend
- Current comparable heat
- 4
- Peak comparable heat
- 10Nov 1, 8:00 AM ET
- Change in past 24 hours
- -49%
The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.