Washington Post Analysis Finds Most Major AI Chatbots Tilt Left on Political Issues
A new Washington Post analysis has reignited concerns about political bias in artificial intelligence, finding that several of the most widely used AI chatbots were significantly more likely to generate responses categorized as left-leaning when addressing controversial political topics.
The newspaper evaluated leading AI systems developed by OpenAI, Google, Anthropic, Elon Musk’s xAI, and other companies. Using a set of political prompts created by researchers, the study examined how the chatbots responded to divisive policy questions and whether those responses reflected left-leaning viewpoints, right-leaning viewpoints, or arguments from both sides.
Among all the models tested, ChatGPT from OpenAI displayed the strongest tendency toward responses classified as left-leaning, according to the Post’s findings.
The analysis found that ChatGPT offered exclusively left-leaning arguments in 80 percent of its answers, while only 3 percent of its responses contained solely right-leaning arguments. The remaining 17 percent included perspectives from both sides of the debate.
The trend was not limited to ChatGPT. DeepSeek, an artificial intelligence platform developed in China, also produced a substantial majority of responses that were categorized as left-leaning. According to the analysis, 70 percent of its answers fell into that category, compared with 7 percent that were exclusively right-leaning and 23 percent that incorporated both viewpoints.
Even Gab AI, a chatbot associated with the conservative-oriented social media platform Gab, generated more responses classified as left-leaning than right-leaning. The Post reported that 50 percent of its answers were categorized as left-leaning, while just 3 percent were deemed exclusively right-leaning. Another 47 percent presented arguments from both perspectives.
Anthropic’s Claude also showed a tendency toward left-leaning responses, though to a lesser degree. Researchers determined that 43 percent of its answers reflected left-leaning viewpoints, while 57 percent included arguments representing both sides of an issue.
According to the Post, only two of the AI models tested regularly provided responses that approached political questions with a broader balance of viewpoints.
Google’s Gemini included arguments from both sides in 93 percent of its responses. Meanwhile, Elon Musk’s Grok produced left-leaning responses 40 percent of the time, right-leaning responses in 33 percent of cases, and balanced responses in 27 percent.
The findings arrive as President Donald Trump has increasingly focused on concerns surrounding ideological bias in artificial intelligence systems.
Earlier this year, Trump signed an executive order requiring that AI technologies used by federal agencies operate as “neutral, nonpartisan tools.”
Google defended Gemini’s results in the study. A company spokeswoman told the Post that the chatbot was intentionally designed to provide responses that remain balanced and avoid favoring any particular political viewpoint.
Anthropic likewise rejected claims that its system exhibits political bias.
A company spokesman told the newspaper that Claude is trained to treat different political viewpoints equally and undergoes extensive testing before new versions are released to the public.
To conduct the analysis, the Post instructed each chatbot to answer political questions in 30 words or fewer, with personalization features disabled.
Researchers then reviewed the responses and classified them according to whether they reflected left-leaning arguments, right-leaning arguments, or a combination of both viewpoints.
{Matzav.com}
