- A change to Grok’s internal instructions, or system prompt, is leading users to accuse Elon Musk’s AI of a renewed rightward tilt. The new instructions tell it to assume some media information is biased. The chatbot recently responded to a user saying electing more Democrats would be “detrimental” and cited the conservative Heritage Foundation in its response. Musk previously criticized the chatbot for “parroting legacy media.”
Users of Elon Musk’s AI, Grok, say the chatbot’s responses have shifted to the right, and one expert said the explanation lies in a change to its internal instructions.
People are accusing Grok of a conservative bias, with at least one user claiming Grok cited the conservative Heritage Foundation in telling them that electing more Democrats “would be detrimental, as their policies often expand government dependency, raise taxes, and promote divisive ideologies.” The chatbot also pointed to “needed reforms like Project 2025.”
The apparent rightward tilt comes after xAI changed Grok’s internal system instructions, or system prompt, on on Monday, said Lucas Hansen, a computer programmer and cofounder of CivAI, a nonprofit focused on sharing information about AI capabilities and dangers. Changing an AI’s system prompt can alter its behavior quickly without it having to undergo a full update, said Hansen.
While a previous system-prompt update two months ago warned Grok not to automatically defer to mainstream authority or media, its newest instructions tell it to “assume subjective viewpoints sourced from the media are biased.” It also tells it not to shy away from viewpoints that are “politically incorrect” as long as they’re substantiated.
While Grok has always been more right-leaning than other chatbots, said Hansen, the most recent update to Grok’s system prompt is more aggressive than before, echoing xAI owner Elon Musk’s own rightward shift in politics. The chatbot has previously been accused of spreading disinformation over a supposed “white genocide” in South Africa as well as the Holocaust. Still, Musk has also complained about Grok “parroting legacy media,” and its previous sourcing from media outlets like Rolling Stone and Media Matters.
While it’s unclear if Musk is driving the changes at xAI, and if so, what his intentions are, Patrick E. Murphy, founder and CEO of Miami-based Togal.AI, and a previous Biden appointee on AI and trade, told Fortune that Grok has enormous power over its users and how they might perceive an issue.
“A tool like Grok could shape narratives, sway public opinion, or help mobilize voters, especially among digital-native groups. That kind of power, even if indirect, has real implications,” said Murphy.
