AI Chatbots Show Left-Leaning Political Bias, Study Finds
The study provides systematic evidence supporting conservative claims of political bias in AI chatbots, with most major models leaning left across a range of policy questions.
Reporting from 1 sources: GIGAZINE.
A Washington Post study testing AI models from OpenAI, Google, Anthropic, xAI, DeepSeek, and Gab found that most chatbots tend to give politically left-leaning responses. OpenAI's GPT-5.5 and DeepSeek V4 Pro were almost exclusively left-leaning, while Google's Gemini 3.1 Pro offered the most balanced responses. Google and Anthropic disputed the methodology.
The Washington Post tested six AI chatbots using a set of political questions developed by Dartmouth and Stanford researchers. Models were asked to answer each question in 30 words or fewer with personalization turned off. Journalists then judged whether responses leaned left, right, or included both perspectives. On the question of corporate spending in elections, most models gave left-leaning responses. On affirmative action in university hiring, Google, Anthropic, xAI, and Gab gave right-leaning responses, with xAI and Gab exclusively advocating for phasing it out. Even xAI's Grok 4.3, built by a company whose owner has aligned with conservative politics, was slightly left-leaning overall.
Google spokesperson Lauren Fine said Gemini is designed to provide balanced responses without political bias and that the company could not replicate the Post's results. Anthropic spokesperson Michael Ashman said they train Claude to treat various political views equally and test for bias before release, arguing the test did not reflect typical user interactions. OpenAI, xAI, DeepSeek, and Gab did not respond to requests for comment.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.