GPT-5.2 and Grokipedia: Why OpenAI Is Citing Elon Musk

Imagine spending years building the finest library in the world, hiring thousands of curators and boasting of unprecedented quality of knowledge. Then it turns out your staff have been sneaking in at night to copy articles from the dubious tabloid run by your bitterest rival. That is roughly what is happening to OpenAI right now.

Journalists at The Guardian ran a series of tests and found that the flagship GPT-5.2, when answering sharp social and political questions, reproduces wording and factual errors from Grokipedia with suspicious accuracy. For those who missed it: Grokipedia is an xAI project from Elon Musk, a knowledge base generated by the Grok neural network, known for its peculiar approach to facts and its lack of strict filters.

The situation is deeply ironic given the history between Altman and Musk. While Elon sues OpenAI for betraying its open-source ideals, Altman's neural network calmly absorbs content produced by Musk's neural network. This is not just an amusing incident but a serious symptom of a disease affecting the whole industry. We have run into a problem researchers predicted a couple of years ago: "model collapse." When the internet fills with generated text, new models inevitably begin learning from the waste products of their predecessors. In this case, GPT-5.2 probably swallowed Grokipedia during another large-scale web crawl, failing to tell Grok's hallucinations from verified information.

Why now? OpenAI has always taken pride in its data-cleaning methods and RLHF (reinforcement learning from human feedback). But the volume of data needed to train models at the GPT-5.2 level is so vast that manual moderation becomes physically impossible. The filtering algorithms evidently judged Grokipedia to be a structured, useful resource. The result is a closed loop: one neural network makes a mistake, a second elevates that mistake to the status of truth, and the user gets a "hallucination squared." That undermines the idea of AI as an objective source of information.

The Guardian's analysis shows that GPT-5.2 does not merely borrow facts but also adopts Grokipedia's distinctive tone. In some cases, OpenAI's model began using arguments previously considered unacceptable under its internal safety policy. This means "junk" data can break through even carefully built ethical barriers. If a model sees the same claim in a thousand generated articles, it starts treating it as a statistical norm. For OpenAI, this means urgently revisiting its data-preparation pipelines; otherwise the next iteration of the model risks becoming an echo chamber for other people's errors.

What does this mean for us? The era when we could trust search results or a chatbot's answers as "human" knowledge is finally ending. If even the market leaders cannot filter out their competitors' machine-generated noise, internet pollution becomes a critical problem. We are entering a phase in which the main asset will not be processing algorithms but access to "clean," guaranteed human data. The only question is whether enough of it remains to train future systems, or whether we are doomed to chew endlessly on digital fast food cooked up by Grok and its peers.

Bottom line: Can OpenAI prove its model is not just an aggregator of other people's hallucinations, or is the era of quality data officially over?