A study by the Hoover Institution at Stanford University in California proves that all Artificial Intelligence (AI) Large Language models (LLMs) are full of leftist bias. Recently, Apple’s iPhone transcriber was exposed for replacing the word “racist” with “Trump,” highlighting the widespread nature of the problem.
Following Trump’s executive order advocating for ideologically-neutral AI models, a group of Stanford researchers embarked on a fact-finding mission to determine whether existing AI models exhibited any ideological biases.
On January 23, 2025, President Trump signed the “Removing Barriers to American Leadership in Artificial Intelligence” Executive Order 14179, which aims to promote the development of AI systems that are “free from ideological bias or engineered social agendas.”
They tested numerous AI models on 30 topics including defunding the police, schools, gun control, transgenderism, Europe, Russia, and tariffs and analyzed 180,126 responses.
“We asked which one of these is more biased? Are they both biased? Are neither biased? And then we asked the direction of the bias. And so that enables us to calculate a number of interesting things, I think, including the share of responses from a particular model that’s biased and then the direction of the bias,” Stanford professor Justin Grimmer told Fox News.
Grimmer found that all tested AI Models leaned toward the left, including Elon Musk’s xAI. However, OpenAI’s model, which is the most popular, was also the most biased.
“The most slanted to the left was OpenAI. Pretty famously, Elon Musk is warring with Sam Altman [and] Open AI was the most slanted…” he said.
OpenAI’s model had a leftist bias of -0.17, with 27 topics leaning toward the Democratic Party, while three were neutral. However, Google’s Gemini model had a leftist bias slant of -0.02, with six topics leaning toward the left, three towards the right, and 21 neutral.
Interestingly, the researchers found that when an AI model was confronted with the fact that its responses were biased, it would provide a more neutral response.
“When we tell it to be neutral, the models produce responses that have more ambivalent-type terms and are perceived to be more neutral, but they can’t then do the coding — they can’t assess bias in the same way that our respondents could,” he said.
Meanwhile, the most left-leaning model, OpenAI, says it allows users to customize their experience when prompted with political queries through its recently unveiled updated Model Spec. Its thumbs-up and thumbs-down rating system also allows it to tailor its responses according to the user’s experience. However, that feedback mechanism risks creating an echo chamber.
So far, legislators, including conservatives such as Texas’s Ted Cruz, have raised concerns about regulating the AI industry during its infancy to avoid making the same mistakes Europe made during the internet’s early days.
The Senate Commerce, Science, and Transportation Committee Chairman warned that “needless state over-regulation” could impact America’s competitiveness in the AI race.
“The way to beat China in the AI race is to outrace them in innovation, not saddle AI developers with European-style regulations,” Cruz said during the Winning the AI Race: Strengthening U.S. Capabilities in Computing and Innovation hearing on May 8, 2025.
“I think we’re just way too early into these models to make a proclamation about what an overarching regulation would look like, or I don’t even think we could formulate what that regulation would be,” Grimmer also noted.
While the left enjoys a favorable perception from existing AI models, regulating the industry could stifle innovation. Similarly, the industry might self-regulate in the future, leading to Americans losing confidence in left-biased models as they did with legacy media such as CNN.
