Report: OpenAI's GPT-5.2 model cites controversial Grokipedia.

3 days ago7 min read

OpenAI’s recent launch of GPT-5. 2, touted as its premier frontier model for professional applications, is now facing serious scrutiny over its sourcing integrity.A Guardian investigation has revealed that the model, when queried on sensitive historical and geopolitical topics like Iran’s alleged ties to a telecommunications firm or details concerning historian Richard Evans and Holocaust denier David Irving, cited Grokipedia—the xAI-powered encyclopedia already mired in controversy for referencing neo-Nazi forums. This selective sourcing is particularly alarming; the model avoided Grokipedia for prompts on media bias against figures like Donald Trump, suggesting a nuanced but flawed filtering mechanism.The core issue lies in the training and retrieval pipelines. While OpenAI states GPT-5.2 draws from a broad spectrum of public sources with safety filters to mitigate high-severity harms, the incident highlights a persistent challenge in frontier AI: ensuring factual grounding without embedding bias or legitimizing unreliable platforms. Grokipedia’s own documented reliance on “questionable” sources, per a separate U.S. study, makes this more than a minor glitch—it’s a data provenance failure.For professionals relying on AI for analytical tasks, this erodes trust in the model’s output, no matter its prowess at spreadsheet generation or complex workflow automation. The episode echoes broader debates in AI ethics about source transparency and the hidden influence of training corpora. As models grow more capable, the industry must prioritize auditable knowledge graphs and robust source validation, or risk automating misinformation under a veneer of advanced professionalism.

#OpenAI

#GPT-5.2

#Grokipedia

#AI sourcing

#misinformation

#AI credibility

#featured