arXiv, the foundational preprint server that has hosted nearly 3 million research papers since its 1991 founding, is becoming an independent nonprofit organization on July 1, 2026. The move comes after more than two decades as part of Cornell University, driven by mounting operational costs and an overwhelming surge in AI-generated submissions.
The platform's annual operating costs have reached approximately $6.7 million, with a reported deficit of nearly $300,000 in 2025. Submissions increased 17% year-over-year in 2025, with rejection rates climbing to between 4% and 12% due to a surge in low-quality and fraudulent AI-generated research—commonly referred to as "AI slop."
"Universities, for all their virtues, have no history of providing this sort of global research infrastructure over a long term," arXiv founder Paul Ginsparg said in a statement.
Governance and Financial Structure
The independent nonprofit will operate under professional leadership rather than academic volunteer management. An international search is underway for arXiv's first chief executive officer, with an advertised salary of approximately $300,000. Governance will shift to a Board of Directors representing the international scientific community, including representation from Cornell and the Simons Foundation.
Cornell and the Simons Foundation have raised startup capital to ensure arXiv begins independence without a deficit and with funding for multiple years of operation. The platform plans no immediate increase in membership fees.
AI Slop Crisis
The transition will enable faster technological development and improved moderation systems to combat AI-generated submissions. arXiv's approximately 300 specialized volunteer moderators handle submissions across numerous subcategories, but this is insufficient for global daily volume. A January 2026 blog post announced a new English-only policy for papers to reduce moderation costs, as recruiting language-specific moderators is unfeasible.
The research community has responded enthusiastically—the independent arXiv dataset has already received over 2,500 downloads in its first week on Hugging Face.