GPTZero finds 100 new hallucinations in NeurIPS 2025 accepted papers - Hacker News

A recent analysis by GPTZero has uncovered over 100 instances of AI-generated inaccuracies, referred to as hallucinations, in papers accepted at the NeurIPS 2025 conference. This revelation highlights significant concerns regarding the integrity of academic research in artificial intelligence, as these fabricated citations managed to pass through the peer review process. The findings were reported across various platforms, including Hacker News and Fortune, drawing attention to the potential implications for the credibility of AI research. The persistence of this issue underscores a broader trend in the academic landscape, where reliance on AI tools for research and writing may compromise the quality and reliability of scholarly work. This phenomenon reflects historical challenges in academia, such as the prevalence of plagiarism and the difficulties in maintaining rigorous standards in peer review. As AI technologies continue to evolve and integrate into research practices, the risk of misinformation and the dilution of academic rigor may become increasingly pronounced. This situation emphasizes the need for enhanced scrutiny and validation processes in academic publishing, as the integrity of research is crucial for advancing knowledge and innovation in the field of artificial intelligence.