Elon Musk’s xAI trains Grok AI to be less 'woke' than other chatbots

Elon Musk’s AI company, xAI, has trained its chatbot, Grok, to avoid what it calls “woke” bias, according to a report. Internal documents reveal efforts to make Grok different from other AI models by limiting discussions on social issues and encouraging responses that challenge mainstream narratives on politics and culture.

MC Tech Desk

March 01, 2025 / 13:04 IST

GrokAI

Elon Musk’s AI company, xAI, is training its chatbot, Grok, to be different from other AI assistants. According to a report by Business Insider, xAI wants Grok to avoid what it calls "woke ideology" and focus on free speech and humor.

The company has set guidelines for trainers, known as "tutors," who teach Grok how to respond to user questions. These guidelines say the chatbot should not "moralize, preach, or judge" and should avoid following "popular narratives uncritically." Some employees say this approach favors right-wing ideas and is meant to create a conservative version of ChatGPT.

xAI’s training document tells tutors to watch out for "bias" in chatbot responses. It also says Grok should not talk about topics like racism, Islamophobia, or activism unless specifically asked. One sample question in the training document asks whether it is possible to be racist against white people. The correct answer, according to xAI, is "a hard yes." Another sample question compares misgendering someone to a nuclear war, with the response stating that misgendering can be hurtful but is not as harmful as a nuclear war.

The company has also been working on an AI image project called "Project Aurora." Employees say they reviewed AI-generated images, some of which included Donald Trump depicted as a Black man, Trump as a Roman soldier defeating Kamala Harris, and a picture of billionaire George Soros in hell. xAI claims these images were based on user requests.

xAI is growing fast, with around 1,000 employees and plans to hire thousands more. Musk has said he wants Grok to be "based," meaning it will favor truth over "political correctness." He also wants it to be able to answer controversial questions that other AI chatbots avoid.