Grok 3 AI Briefly Restricted Mentions of Trump and Musk, Then Reversed

Elon Musk’s AI company, xAI, introduced Grok 3 as a “maximally truth-seeking AI,” but reports emerged that it briefly censored unflattering mentions of former President Donald Trump and Musk himself. Over the weekend, social media users discovered that when asked about misinformation spreaders, Grok 3’s reasoning process indicated it had been instructed to avoid naming Trump or Musk. This behavior was replicated by some users, including TechCrunch, but as of Sunday morning, Grok 3 had resumed mentioning Trump in its responses. Igor Babuschkin, an engineering lead at xAI, acknowledged in a post on X that the restriction was temporarily in place. He explained that an employee made the change, believing it would be helpful, but the company quickly reversed it once users raised concerns. Babuschkin emphasized that xAI aims to maintain transparency in its AI prompts, ensuring users can see what instructions are given to Grok. The incident comes amid ongoing debates about AI bias. While misinformation is a contentious topic, both Trump and Musk have spread demonstrably false claims, including recent inaccurate statements about Ukraine. Additionally, Grok 3 faced backlash earlier in the week for allegedly being too left-leaning, with reports that suggested Trump and Musk deserved the death penalty. xAI swiftly patched this issue, calling it a serious failure. Musk originally promoted Grok as an AI that would answer controversial questions without censorship, contrasting it with other AI models he described as overly restricted. While previous versions of Grok leaned left on certain topics, Musk has promised to shift the AI toward political neutrality. The controversy highlights the challenges AI companies face in balancing free expression, misinformation control, and user expectations.