How AI slop is causing a crisis in computer science

Nature

by Elizabeth Gibney

February 14, 2026

AI-Generated Deep Dive Summary

AI slop—low-quality or entirely AI-generated research papers—is overwhelming computer science journals and conferences, creating a critical crisis in scientific publishing. With tools like OpenAI's Prism enabling researchers to draft papers in just minutes without conducting actual experiments, the flood of submissions has surged beyond manageable levels. For instance, the 2026 International Conference on Machine Learning (ICML) received over 24,000 submissions, more than double its previous year’s count. This influx, combined with a lack of proper validation by authors and widespread AI-generated fabrications known as "hallucinations," has left peer reviewers overwhelmed and struggling to maintain rigorous evaluations. The problem extends beyond sheer volume. Many submissions are entirely AI-generated or contain fabricated content that is difficult to detect through conventional methods like abstracts or skimming texts. This poses an existential threat to the scientific publishing system, as volunteer moderators and reviewers can no longer reliably assess the quality or validity of these papers using traditional criteria. To combat this issue, preprint repositories and conference organizers are implementing new measures. arXiv has introduced eligibility checks for first-time submitters and banned certain types of review articles. The International Joint Conferences on Artificial Intelligence (IJCAI) now charges $100 per additional submission, with proceeds going to reviewers. Other conferences have expanded their reviewer pools or incentivized high-quality reviews through waived registration fees. The stakes are high: if the scientific community fails to address this crisis, trust in research could erode, particularly within computer science. The rapid adoption of AI tools has not only accelerated research productivity but also made it easier to produce fake or low-quality work, threatening the integrity of scientific publishing and collaboration. Addressing this challenge is crucial for maintaining the credibility and reliability of scientific advancements. This issue matters deeply to anyone interested in the future of science and technology. If left unchecked, AI slop could undermine decades of progress by eroding trust in research quality and integrity. The scientific community

Verticals

scienceresearch

Originally published on Nature on 2/14/2026