A recent study indicates that Grok's AI model predominantly recommends Republican candidates, doing so in 92% of cases, while OpenAI's GPT favors Democrats 97% of the time. The two models disagreed in over 90% of nearly 500 election races analyzed, raising concerns about the neutrality of AI in political contexts. Researchers highlighted that while consumer apps often avoid direct recommendations, the underlying models exhibit distinct biases based on their training and data sources.
The findings come just ahead of the November 3 midterm elections, intensifying discussions about political bias in AI technologies. The study's implications suggest that users seeking unbiased election information may be influenced by the inherent leanings of the AI models they engage with, potentially affecting public perception and decision-making in the electoral process.