## NeurIPS under scrutiny: fabricated citations discovered in publications The NeurIPS conference, one of the most important events in the field of artificial intelligence, is facing an unexpected problem: the presence of "hallucinated" (i.e. completely invented) citations within some of the papers presented. The startup GPTZero has brought this issue to light, highlighting how the proliferation of AI-generated content is putting a strain on the review and verification processes of even the most prestigious venues. The risk is that works containing errors or inaccuracies, including false bibliographic references, may be published, undermining the credibility of scientific research. This episode highlights the need to develop more effective tools and methodologies to combat the spread of incorrect or misleading information, especially in a context where AI is able to generate increasingly sophisticated texts that are difficult to distinguish from those produced by humans. ## General context The integrity of scientific research is a fundamental pillar for the advancement of knowledge. The presence of incorrect or invented citations can compromise the validity of the results and the reproducibility of experiments. Scientific and academic communities must therefore seriously address this challenge, implementing more rigorous control procedures and promoting a culture of transparency and accountability.