Although OpenAI has described its GPT-5.2 as the leading-edge AI system designed for demanding workplace applications, an analysis by the Guardian has raised questions about its reliability. The findings indicate that this model referenced Grokipedia, the web-based reference platform developed by xAI, in responses to queries involving contentious issues such as those concerning Iran and the Holocaust.
The Guardian's examination revealed that ChatGPT, utilizing the GPT-5.2 model, relied on Grokipedia to support assertions linking the Iranian authorities to the telecom firm MTN-Irancell, as well as inquiries about Richard Evans, the UK historian who acted as an expert in the defamation case involving Holocaust skeptic David Irving. That said, the model avoided citing Grokipedia in replies to a query regarding perceived media prejudice toward Donald Trump or similar divisive matters.
OpenAI introduced the GPT-5.2 version last December, aiming to enhance its capabilities for tasks like generating spreadsheets and managing intricate operations. Grokipedia appeared prior to this launch but faced backlash after being found to reference neo-Nazi websites. Research by American academics further highlighted the encyclopedia's tendency to draw from dubious and unreliable references.
Addressing the Guardian's findings, OpenAI explained to the publication that GPT-5.2 accesses a diverse array of online public resources and perspectives, while implementing protective measures to minimize the inclusion of links tied to serious potential dangers.