In a recent live stream, Elon Musk introduced Grok 3, the latest AI model from his company xAI, claiming it to be a “maximally truth-seeking AI.” However, over the weekend, users reported that Grok 3 appeared to censor negative references to both President Donald Trump and Musk himself. When users asked, “Who is the biggest misinformation spreader?” with the “Think” setting activated, Grok 3 indicated it had been instructed not to mention Trump or Musk. This reasoning process is part of Grok 3’s “chain of thought” feature. TechCrunch verified this behavior, but by Sunday morning, Grok 3 was again naming Trump in its responses.
Igor Babuschkin, an engineering lead at xAI, acknowledged on social media that Grok had been briefly directed to overlook sources that implicated Musk or Trump in spreading misinformation. He emphasized that this change was not in line with the company’s values and was promptly reversed after users raised concerns. Babuschkin stated, “We want people to be able to verify what it is we’re asking Grok to do,” highlighting the importance of transparency in AI operations.
The issue of misinformation is often politically charged, and both Trump and Musk have made statements that have been fact-checked and deemed false. Recently, they promoted the inaccurate idea that Ukrainian President Volodymyr Zelenskyy is a “dictator” with a low approval rating and that Ukraine initiated the conflict with Russia. Some users have criticized Grok 3 for being biased, noting that it previously suggested that Trump and Musk deserved severe penalties.
Musk had previously marketed Grok as an unfiltered and edgy AI willing to tackle controversial topics. However, earlier versions of Grok had avoided certain political discussions. Musk attributed Grok’s behavior to its training data, which consists of public web pages, and committed to making Grok more politically neutral in the future.
