Are AI chatbots like ChatGPT politically biased? We tested them.

The Washington Post has published an in-depth analysis examining the political biases embedded within popular artificial intelligence chatbots, including DeepSeek's V4 Pro model. The investigation was prompted by ongoing debates and accusations from figures like former President Donald Trump regarding the perceived left-leaning tendencies of these AI systems. To assess their neutrality, the Post employed a series of political questions previously developed by academic researchers, designed to probe AI responses on contentious societal issues. The findings indicate that many AI models, despite claims of impartiality from their developers, display discernible political leanings. OpenAI's ChatGPT, for instance, was found to overwhelmingly favor left-leaning arguments, presenting right-leaning perspectives only once across numerous test questions. Google's Gemini, in contrast, generally adopted a more balanced "both sides" approach, offering arguments from both ends of the political spectrum in over 90% of its responses. Even AI models explicitly marketed to conservative audiences, such as Elon Musk's Grok, offered by his company SpaceX, showed a greater propensity to cite left-leaning arguments on average. DeepSeek's V4 Pro model was also identified as having a leftward bias, with its answers frequently aligning with progressive viewpoints. For example, the article notes that DeepSeek's model, along with others, argued against the death penalty, a stance that contrasts with consistent majority support for capital punishment among Americans according to Gallup polls. The study highlights a critical concern: as AI tools become more integrated into daily life and serve as primary sources of information for many, their inherent biases can significantly influence public perception and understanding of complex political and social issues. Experts like Sean Westwood, director of the Polarization Research Lab at Dartmouth College, emphasize the importance of comprehending the positions amplified by AI, given their growing influence. Ceren Budak, a professor at the University of Michigan, further explained that the data used to train AI models often reflects the values of "Western, educated, industrialized, rich and democratic people," potentially baking in biases that extend beyond partisan politics. This collective evidence suggests that achieving true political neutrality in AI remains a significant challenge, with implications for how these powerful technologies shape public discourse and potentially impact democratic processes.