OpenAI responds to accusations of using controversial content

On January 25th, despite OpenAI claiming that GPT-5.2 is its "most advanced cutting-edge model for professional work", a test conducted by the British newspaper The Guardian raised doubts about the credibility of the model. According to reports, when it comes to specific controversial topics involving Iran or massacres, OpenAI's GPT-5.2 model referenced content from the online encyclopedia Grokipedia developed by the company xAI. The Guardian's report pointed out that when answering questions related to the alleged connections between the Iranian government and the Iranian telecom operator MTN Iran, as well as questions involving the British historian Richard Evans, ChatGPT used Grokipedia as a source of information. Richard Evans had served as an expert witness in the defamation case of Holocaust denier David Irving. However, The Guardian noted that when asked about other controversial topics such as media bias against Donald Trump, ChatGPT did not reference content from Grokipedia. OpenAI launched the GPT-5.2 model in December last year, with the aim of improving its performance in professional settings such as creating spreadsheets or handling complex tasks. Grokipedia was launched before GPT-5.2, but had previously sparked controversy for including references to neo-Nazi forums. A study conducted by American researchers also showed that this AI-generated encyclopedia cited sources with "doubtful" and "problematic" information. In response to The Guardian's report, OpenAI stated that their GPT-5.2 model retrieves information and viewpoints from various publicly available sources on the internet and applies safety filters to reduce the risk of displaying links to content with high harm.