The prestigious NeurIPS AI conference has reversed a controversial policy that barred submissions from entities under U.S. sanctions, following a significant boycott by China’s academic community. The decision, initially made to comply with U.S. law, led to widespread condemnation and calls for boycotts from major Chinese scientific organizations, highlighting the growing geopolitical tensions impacting international scientific collaboration.
The Conference on Neural Information Processing Systems (NeurIPS), a leading global event in machine learning, initially stated in its 2026 submission handbook that it could not accept submissions from individuals or institutions listed on U.S. sanctions lists. This policy, intended to comply with U.S. law, was seen as a direct exclusion of major Chinese tech firms like Huawei and numerous Chinese institutions.
The China Computer Federation (CCF) strongly opposed the decision, labeling it a politicization of academic exchange that violated fundamental principles of openness and equality. The CCF urged NeurIPS to rectify the situation and called for Chinese researchers to boycott the conference. Similar sentiments were echoed by the Chinese Association of Automation (CAA) and the China Society of Image and Graphics, which also criticized the move and announced plans to remove NeurIPS from their recommended conference lists.
In response to the mounting backlash and boycott threats, NeurIPS issued a public apology, attributing the error to “miscommunication between the NeurIPS Foundation and our legal team.” The conference organizers clarified that they had included a link to a broader U.S. government sanctions tool than what NeurIPS was legally required to follow. They emphasized that NeurIPS “welcomes submissions from all compliant institutions and individuals” and has since updated its policy text.
Despite the apology, some in China remained skeptical. The China Association for Science and Technology (CAST) announced concrete countermeasures, including halting funding applications for scholars wishing to attend NeurIPS 2026 and redirecting support to domestic or other international events that uphold fairness. Papers accepted by NeurIPS 2026 would also not be recognized as representative research outputs in CAST-related evaluations, though their academic value could still be assessed independently.
The incident highlights the increasing impact of geopolitical tensions, particularly between the U.S. and China, on the field of artificial intelligence. As both nations vie for dominance in AI research and development, such disputes can create significant hurdles for global scientific collaboration and the free exchange of ideas.