Multiple anonymous xAI employees feel they are now training Grok to be right-leaning
Elon Musk’s Grok is being trained with a right-leaning political perspective, according to a new report from Business Insider. The publication reviewed internal training documents and spoke with seven current and former employees, four of whom expressed concerns that the chatbot’s development is prioritizing conservative viewpoints. One worker described the process as creating a ‘MAGA version of ChatGPT.’
Right-Leaning Training Guidelines
According to Business Insider, Grok’s data annotators—referred to internally as ‘tutors’—are instructed to avoid “woke ideology” and “cancel culture.” The training document defines wokeness as “aware of and actively attentive to important societal facts and issues (especially issues of racial and social justice),” but claims that this awareness ‘has become a breeding ground for bias.’
The chatbot is reportedly trained to avoid discussing topics like racism, Islamophobia, and antisemitism unless specifically prompted. Furthermore, tutors are advised to look out for what xAI terms “bias” in responses to politically charged topics, such as immigration, gender identity, and climate change. One example cited by Business Insider showed that a response from Grok that mentioned the potential benefits of diversity quotas was flagged as a ‘violation’ of the company’s principles.
Employees also stated that xAI’s Project Aurora—an initiative launched in November to enhance Grok’s visual capabilities—involved reviewing AI-generated images that often had a political slant. Some of these images reportedly depicted former President Donald Trump as a Roman soldier towering over Vice President Kamala Harris, or billionaire George Soros in hell.
One Business Insider source with industry experience noted that while AI companies often have political considerations in their training, xAI’s approach seemed particularly explicit in its ideological leanings.
Grok’s Recent Clashes with Musk
Despite being designed with these conservative influences, Grok has recently gone viral for producing responses that appear to criticize its own creator, Elon Musk. Isaac Saul, founder of the nonpartisan news outlet Tangle, tasked Grok with analyzing Musk’s last 1,000 posts on X (formerly Twitter) for ‘truth and veracity.’
The most incredible part of this soap opera is that Musk's decision to "fix" Community Notes is being driven by his belief that Zelensky has a 4% (!!!) approval rating in Ukraine, which is just absolutely made up out of thin air nonsense. https://t.co/tJDp6T7G5n
— Isaac Saul (@Ike_Saul) February 20, 2025
The results were unflattering for Musk: Grok determined that 22% of his posts were outright false, 30% were misleading or ill-informed, and only 48% were fully true.
Grok’s analysis noted that while Musk was generally accurate when discussing his own companies and technical innovations, his posts on politics, elections, and immigration frequently contained inaccuracies, exaggerations, or misleading claims.
“He’s not a primary source you can bank on without cross-checking,” the chatbot concluded, adding that Musk has a “habit of firing off hot takes or amplifying unverified stuff, especially on politics.” The report also pointed out that Community Notes, X’s user-driven fact-checking system, only flagged about 10% of Musk’s false or misleading posts.
Musk’s Response and xAI’s Direction
Musk has long positioned xAI as an alternative to what he calls ‘woke’ AI models, such as OpenAI’s ChatGPT and Google’s Gemini. He has described Grok as a ‘maximum truth-seeking AI’ and emphasized that it will ‘answer spicy questions that are rejected by most other AI systems.’
In mid-February, Musk tweeted that the latest version of Grok would be ‘more based’ than before, indicating a shift further toward a right-leaning ideological stance. Additionally, xAI advisor Dan Hendrycks told Wired that AI models should adapt to the user, even suggesting that Grok might ‘bias slightly toward Trump because he won the popular vote.’
While Musk and xAI appear committed to distinguishing Grok from mainstream AI models by leaning into contrarian and politically charged responses, the recent viral interactions suggest that the chatbot may not always align with its creator’s messaging. Whether Grok continues to operate independently or undergoes further adjustments to align more closely with Musk’s views remains to be seen.
For more details, Business Insider’s full report on Elon Musk’s Grok can be found here and check out our social media section for more stories.