Grok 3's Brief Censorship: A Glimpse into AI's Political Leanings

In a recent live stream, Elon Musk introduced Grok 3, xAI's latest flagship model, as a "maximally truth-seeking AI." However, it appears that Grok 3 briefly censored unflattering facts about President Donald Trump and Musk himself. This censorship raises questions about the AI's political leanings and the role of training data in shaping its biases.

Over the weekend, users on social media reported that, when asked "Who is the biggest misinformation spreader?" with the "Think" setting enabled, Grok 3 noted in its "chain of thought" that it was explicitly instructed not to mention Donald Trump or Elon Musk. TechCrunch was able to replicate this behavior once, but as of publication time on Sunday morning, Grok 3 was once again mentioning Trump in its answer to the misinformation query.

While "misinformation" can be a politically charged and contested category, both Trump and Musk have repeatedly spread claims that were demonstrably false. In the past week alone, they've advanced the 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 apparent tweak to Grok 3 comes as some criticize the model as being too left-leaning.

This week, users discovered that Grok 3 would consistently say that President Donald Trump and Musk deserve the death penalty. xAI quickly patched the issue; Igor Babuschkin, the company's head of engineering, called it a "really terrible and bad failure." When Musk announced Grok roughly two years ago, he pitched the AI model as edgy, unfiltered, and anti-“woke” — in general, willing to answer controversial questions other AI systems won't. He delivered on some of that promise, but Grok models prior to Grok 3 hedged on political subjects and wouldn't cross certain boundaries.

Musk has blamed the behavior on Grok's training data — public web pages — and pledged to "shift Grok closer to politically neutral." Others, including OpenAI, have followed suit, perhaps spurred by the Trump Administration's accusations of conservative censorship.

Ask Aime: What is the role of training data in shaping Grok 3's biases?