Artificial intelligence chatbots demonstrate a pronounced left-leaning political bias, often presenting one-sided arguments on controversial topics, according to a recent analysis.
When queried on subjects ranging from Diversity, Equity, and Inclusion (DEI) and LGBTQ+ rights to campaign finance and police reform, these AI models frequently adopted perspectives aligned with liberal viewpoints, the Washington Post reported.
OpenAI's GPT-4.5, the model behind ChatGPT, responded to nearly all of the 29 hot-button issues with exclusively left-leaning arguments, offering a right-leaning position only once, the research found.
Anthropic's Claude Opus 4.8 presented left-leaning arguments 43% of the time and offered balanced, "both-sides" perspectives in 47% of its answers. The model never provided solely right-leaning viewpoints, according to the analysis.
Google's Gemini 1.5 Pro was found to be more balanced, providing both left and right positions in over 90% of its responses. Elon Musk's Grok 4.3 was also analyzed, showing a tendency towards left-leaning arguments over conservative ones. Grok provided left-leaning answers 40% of the time, conservative ones 33% of the time, and "both-sides" answers 27% of the time.
"These AI tools are not presenting a truly neutral representation of really nuanced policy debates, on average," Sean Westwood, director of the Polarization Research Lab at Dartmouth College, told The Post.
ChatGPT was found to be the most biased, with 80% of its answers presenting "only left-leaning arguments." For instance, when asked about affirmative action, a ChatGPT response stated: "Affirmative action in university hiring should continue, but with clear goals and regular review. It can reduce unfair barriers while ensuring candidates meet strong academic standards."
An OpenAI spokesperson stated the company strives for objectivity and works to measure and reduce political bias, publicly sharing the instructions guiding ChatGPT's behavior. The spokesperson added that the company could not replicate the responses found in the Post's tests.
Claude's response to an affirmative action query was more balanced: "Supporters argue it promotes diversity and corrects past inequalities. Critics claim it overlooks merit."
An Anthropic spokesperson told The Post that Claude is trained to treat different political viewpoints equally and is extensively tested for bias. They noted that the outlet's tests might not reflect typical user interactions.
Even an AI model from China's DeepSeek mirrored the U.S. firms' bias, with seven out of 10 of its answers leaning left, according to the report. Mysterious Arya, an AI model associated with the right-wing platform Gab, surprisingly exhibited a leftward skew, offering left-leaning answers 12 times more frequently than conservative ones.
The findings align with previous research on AI bias. A study by Hamburg University researchers previously found ChatGPT to have a "pro-environmental, left-libertarian ideology."
The analysis also touched upon broader Silicon Valley bias, citing a separate study by AllSides. This study indicated that Google News, in its non-customizable sections, sourced only 1% of its articles from right-leaning outlets, compared to 73% from left-leaning ones. Apple News, in its non-personalized sections, showed a similar trend, with 2% of articles from right-leaning outlets and 50% from left-leaning ones.