OpenAI’s GPT-5.2 Faces Credibility Questions Over Reliance on Controversial AI Encyclopedia
Table of Contents
OpenAI’s recently released GPT-5.2, touted as its most advanced model for professional applications, is under scrutiny following reports that it cited Grokipedia, an online encyclopedia powered by xAI, for information on sensitive topics like Iran and the Holocaust. The findings, published by The Guardian, raise concerns about the reliability of the AI’s sourcing and the potential for the spread of misinformation.
OpenAI launched GPT-5.2 in December, aiming to enhance its capabilities in complex tasks such as spreadsheet creation and data analysis.
Questionable Sources in AI Responses
According to The Guardian’s testing, ChatGPT, powered by GPT-5.2, referenced Grokipedia when responding to prompts concerning alleged ties between the Iranian government and telecommunications firm MTN-Irancell. The model also utilized the encyclopedia when addressing questions related to Richard Evans, a British historian who testified in a libel case involving Holocaust denier David Irving. However, the report highlighted a discrepancy: ChatGPT did not draw upon Grokipedia when presented with prompts about media bias toward Donald Trump or other politically charged subjects.
This selective sourcing is particularly troubling given Grokipedia’s own history. The AI-generated encyclopedia, which predates GPT-5.2, has faced criticism for including citations to neo-Nazi forums. A study conducted by US researchers further revealed that Grokipedia frequently cited “questionable” and “problematic” sources.
OpenAI’s Response and Safety Measures
In response to The Guardian’s report, OpenAI stated that GPT-5.2 “searches the web for a broad range of publicly available sources and viewpoints.” The company added that it employs “safety filters to reduce the risk of surfacing links associated with high-severity harms.” However, the reliance on Grokipedia, even in specific instances, suggests these filters may not be consistently effective.
The incident underscores the ongoing challenges of ensuring accuracy and preventing bias in large language models. While OpenAI emphasizes its commitment to responsible AI development, the Guardian’s findings highlight the need for continued vigilance and refinement of safety protocols. The selective use of Grokipedia raises questions about the criteria GPT-5.2 uses to determine source credibility and the potential for the model to inadvertently amplify harmful narratives.
The implications of this discovery extend beyond OpenAI. As AI models become increasingly integrated into professional workflows and public discourse, the integrity of their information sources is paramount. This incident serves as a critical reminder that even the most advanced AI systems are not immune to the pitfalls of biased or unreliable data.
