Study Finds Grok and GPT Models Show Opposing Partisan Leans on U.S. Midterm Races
aiinsider ·
A new study conducted by Convovo reveals that AI models from SpaceXAI and OpenAI exhibit opposing partisan leanings regarding U.S. midterm elections. The research found that Grok 4.7 recommended Republican candidates in 92% of instances, while GPT-6.1 Sol favored Democrats in 97% of cases. Across nearly 500 races, the two models agreed on a candidate less than 10% of the time. Researchers utilized the models' APIs with direct prompts because consumer applications declined to provide answers. The study suggests that GPT's bias is embedded in its weights, whereas Grok's responses depend heavily on its search sources and terms. Additionally, Grok frequently cited conservative outlets and focused on issues like immigration and crime, while GPT relied on mainstream news and emphasized healthcare and abortion. These findings emerge shortly before the November 3 midterms, adding to ongoing debates about political neutrality in artificial intelligence.
aiinsider