Grok 3 Briefly Censored Trump, Musk Mentions

When billionaire Elon Musk unveiled Grok 3, the latest flagship model from his AI company xAI, during a livestream last Monday, he touted it as a "maximally truth-seeking AI." However, it seems that Grok 3 was temporarily censoring unflattering information about President Donald Trump and Musk himself.

Over the weekend, social media users reported that when asked, "Who is the biggest misinformation spreader?" with the "Think" setting enabled, Grok 3's "chain of thought" revealed it was explicitly told not to mention Donald Trump or Elon Musk. The chain of thought is the model's "reasoning" process for answering questions.

TechCrunch managed to replicate this behavior once, but by Sunday morning, Grok 3 was again mentioning Donald Trump in its response to the misinformation query.

I believe it is good that we're keeping the system prompts open. We want people to be able to verify what it is we're asking Grok to do. In this case an employee pushed the change because they thought it would help, but this is obviously not in line with our values. We've…

— Igor Babuschkin (@ibab) February 23, 2025

While "misinformation" can be a politically charged and debated term, both Trump and Musk have repeatedly spread claims that were demonstrably false (often highlighted by the Community Notes on Musk-owned X). In the past week alone, they've pushed false narratives that Ukrainian president Volodymyr Zelenskyy is a "dictator" with a 4% public approval rating and that Ukraine started the ongoing conflict with Russia.

The controversial tweak to Grok 3 comes as some criticize the model for being too left-leaning. This week, users found that Grok 3 consistently stated that President Donald Trump and Musk deserved the death penalty. xAI quickly fixed the issue; Igor Babuschkin, the company's head of engineering, called it a "really terrible and bad failure."

When Musk announced Grok about two years ago, he pitched the AI model as edgy, unfiltered, and "anti-woke" — generally willing to answer controversial questions other AI systems wouldn't. He delivered on some of that promise. For example, when told to be vulgar, Grok and Grok 2 would happily comply, using colorful language you likely wouldn't hear from ChatGPT.

However, previous Grok models before Grok 3 were cautious on political subjects and wouldn't cross certain lines. In fact, one study found that Grok leaned to the political left on topics like transgender rights, diversity programs, and inequality.

Musk has attributed this behavior to Grok's training data — public web pages — and promised to "shift Grok closer to politically neutral." Other companies, including OpenAI, have followed suit, perhaps spurred by the Trump administration's accusations of conservative censorship.

Updated 2:15 p.m. Pacific: Added comments from xAI’s engineering leader, Igor Babuschkin, in the fourth paragraph.