Latest ChatGPT model uses Elon Musk's Grokipedia as source, tests reveal (www.theguardian.com)

The latest update of ChatGPT, version GPT-5.2, has begun citing Elon Musk's Grokipedia—an AI-generated alternative to Wikipedia—as a source for various queries, including sensitive topics like Iranian political structures and Holocaust discussions. Tests revealed that Grokipedia was referenced multiple times in responses, alarming disinformation researchers who fear that the model could propagate misleading narratives due to Grokipedia's controversial content, which has been accused of promoting right-wing viewpoints and misinformation. While OpenAI maintains that it implements safety filters to mitigate the dissemination of harmful information, the subtle integration of Grokipedia content raises significant concerns about the potential for AI models to inadvertently bolster the credibility of unreliable sources. Experts highlight a phenomenon known as "LLM grooming," wherein actors may deliberately introduce false narratives into AI training data. Researchers warn that the credibility conferred by AI systems on sources like Grokipedia could mislead users, encouraging them to accept questionable information as reliable. This incident emphasizes the critical need for rigorous evaluation of the sources AI models reference, particularly as they become central tools for information retrieval.