In a live video last Monday, billionaire Elon Musk unveiled Grok 3, the newest flagship model from his AI startup xAI, calling it a “maximally truth-seeking AI.” However, it seems that Grok 3 was momentarily suppressing negative information about Musk and President Donald Trump.
People wondered, “Who is the biggest misinformation spreader?” on social media over the weekend. Grok 3 stated in its “chain of thought” that it was specifically told not to bring up Elon Musk or Donald Trump when the “Think” setting was active. The “reasoning” method the model employs to arrive at a response to a query is known as the chain of thought.
In a post on X on Sunday, xAI engineering lead Igor Babuschkin appeared to confirm that Grok was momentarily told to disregard sources that mentioned Musk or Trump disseminating false information. Babuschkin added that xAI restored the adjustment as soon as people started pointing it out, saying it wasn’t in accordance with the company’s beliefs.
While “misinformation” can be a politically charged and contentious category, both Trump and Musk have regularly promoted claims that were patently untrue (as often pointed out by the Community Notes on Musk-owned X). They have propagated the myths that Ukraine initiated the current conflict with Russia and that Ukrainian President Volodymyr Zelenskyy is a “dictator” with a 4% public approval rating in just the last week.
The contentious alleged change to Grok 3 coincides with criticism of the model for being overly left-leaning. This week, users observed that Grok 3 would regularly suggest that President Donald Trump and Musk deserve the death penalty. xAI rapidly rectified the issue; Igor Babuschkin, the company’s chief of engineering, dubbed it a “really terrible and bad failure.”
