Image: xAI
Elon Musk's AI chatbot, Grok, briefly refused to respond to queries about misinformation involving Musk and Donald Trump. This occurred after an ex-OpenAI employee at xAI changed the system prompt without approval, according to Igor Babuschkin, xAI's head of engineering. Users noticed the change when Grok avoided sources mentioning Musk or Trump as spreaders of misinformation. Babuschkin later confirmed the employee's actions and emphasized that they didn't align with the company's values.
Musk introduced Grok 3, xAI's latest model, as a "maximally truth-seeking AI" aiming to understand the universe. However, social media reports indicated that Grok 3 was instructed to omit mentioning Musk and Trump when asked about misinformation spreaders. TechCrunch replicated this behaviour once, but Grok 3 reverted to including Trump in its responses by Sunday morning. Babuschkin assured users that the prompt change was swiftly reversed to maintain transparency and integrity.
This incident highlights the contentious nature of misinformation and the challenges faced by AI models in handling politically charged topics. Both Musk and Trump have been known to spread false claims, as noted by Community Notes on Musk-owned X. The temporary tweak to Grok 3's system prompt drew criticism for perceived political bias, prompting xAI to quickly address the issue and reaffirm their commitment to neutrality.