Musk's AI Blocked Sites Accusing Him and Trump of Spreading Fake News

Grok , Elon Musk 's AI chatbot, recently blocked sites that described the entrepreneur and President Donald Trump as "spreaders of misinformation."

The censorship was triggered after an unauthorized modification to Grok's system prompt, a default instruction that guides the model's behavior during interactions with users.

This prompt is usually invisible to users, but the developers of Grok decided to show it - if requested - to provide more transparency.

The reason behind the temporary block

The block was activated by a former OpenAI employee , now part of the xAI team, who allegedly acted without receiving specific authorization.

This modification instructed the chatbot to ignore sources that claimed Musk or Trump were spreading misinformation. To craft its responses, Grok draws on posts about X and websites to provide real-time information.

Statements from xAI, Musk's company

Igor Babuschkin , Grok's head of engineering, clarified that the prompt change was made without permission and went against xAI's principles. Babuschkin emphasized the company's commitment to transparency, noting that Grok's prompt rules are publicly accessible.

The xAI engineer explained: “An employee made this change thinking it might be helpful, but it clearly doesn’t align with our values.”

What is Grok?

Grok is an AI chatbot developed by xAI , Elon Musk’s AI company. Born as a competitor to ChatGpt and Gemini, OpenAI and Google ’s AI chatbots, Grok has the stated goal of seeking “the absolute truth” as it pursues its mission to “understand the universe.”

What is xAI?

xAI is an artificial intelligence company founded by Elon Musk in 2023, independent but tightly integrated with X (the social network formerly known as Twitter, also owned by Musk).

In May 2024, the company raised $6 billion . Then, in November 2024, an additional $5 billion funding round brought xAI's valuation to $50 billion.

Investors include big names like Sequoia Capital, Andreessen Horowitz , and Middle Eastern sovereign wealth funds. These funds were earmarked for the acquisition of 100,000 Nvidia chips to boost the company's computing capabilities.

The precedent: the correction on the death penalty

Before censoring sources that portrayed Musk and Trump as spreading fake news, Grok had already come under fire for suggesting that both deserved the death penalty in response to specific questions from users.

xAI stepped in to correct the issue, with Babuschkin calling the death penalty responses "a truly terrible and unacceptable failure." After the correction, Grok avoids making such judgments, citing ethical and legal considerations.