The handbook arrived quietly on the NeurIPS website on 23 March, a technical update that seemed routine. But the 2026 Main Track Handbook contained a new clause: submissions were now subject to US sanctions compliance. Within hours, the implications became clear. Researchers affiliated with entities on the US Treasury Department's OFAC Specially Designated Nationals list could not submit papers, serve as reviewers, or hold area chair roles. In practical terms, that meant scientists at Huawei, SenseTime, and other major Chinese AI labs were suddenly locked out of what is often billed as the world's premier AI conference.
The response from China's academic establishment was swift and emphatic. On March 25, the China Computer Federation issued a statement condemning the restriction, arguing that barring submissions from certain institutions politicises academic exchange and violates fundamental academic principles. Chinese computer scientists and researchers were urged to boycott the conference after its organisers barred submissions from US-sanctioned institutions, including leading Chinese tech groups such as Huawei Technologies. Within days, researchers from Alibaba and Tencent publicly declined reviewer roles, and senior Tencent researchers resigned from their NeurIPS 2026 positions.
By Friday, 28 March, NeurIPS had capitulated. In a statement, the Conference on Neural Information Processing Systems said it had gone beyond its legal obligations under US sanctions when outlining changes to its submissions policy earlier this week. In preparing the NeurIPS 2026 handbook, it had included a link to a US government sanctions tool that covers a significantly broader set of restrictions than those NeurIPS is actually required to follow, due to miscommunication between the NeurIPS Foundation and its legal team.
Yet the apology arrived too late. CAST announced it would cease funding applications for Chinese members seeking to participate in NeurIPS and would redirect resources toward domestic conferences or international conferences that respect the rights and interests of Chinese academics. The organisation also declared that research published at NeurIPS would no longer count as qualifying output for its funding programmes.
The question gnawing at observers is whether this episode represents mere bureaucratic bungling or something more systematic. The precedent is instructive: in 2019, IEEE barred Huawei researchers from peer review citing similar OFAC compliance concerns, the CCF threatened a boycott, and IEEE reversed its decision within weeks, clarifying that peer review was not a service under sanctions rules. NeurIPS's initial framing suggested a fundamentally different legal assessment in 2026.
But here is where the tension becomes visible. Many Chinese observers saw the timing as arbitrary, noting that US sanctions on these companies have been in place for years, yet other major conferences have not implemented similar submission restrictions. That observation cuts both ways. If other conferences had long refrained from such explicit bans, why did NeurIPS suddenly feel compelled to act? And if the legal environment has genuinely tightened, why was the policy so quickly reversed?
The dispute matters because it ties AI research access to geopolitics: major AI conferences typically function as neutral marketplaces for ideas, paper review, and collaborations. When participation rules explicitly track sanctions, conferences shift from technical gatekeepers toward political fault lines.
What makes this pause, not so quietly, is what comes next. Without careful management, the AI field could become bifurcated along geopolitical lines, with China and the United States developing partially segregated research ecosystems. Whether the conference organisers' statement that they have updated the link and clarified the text to be consistent with that of other international conferences and NeurIPS in the past represents genuine clarification or renewed vagueness remains, from the Chinese perspective, genuinely unclear.
The real damage may already be done. Some observers argue that the swift apology from NeurIPS suggests a sense of panic, suggesting that without Chinese researchers contributing top-tier intellectual talent and robust data, the conference would struggle to maintain its status as a leading international venue. That observation, whether accurate or not, has already poisoned the well. Trust, once fractured in academic collaboration, takes years to rebuild.
The conference is scheduled for December in Sydney. Whether enough Chinese researchers return to salvage its international character, or whether the boycott hardens into something permanent, remains the open question. What is no longer in question is whether science can remain insulated from geopolitical competition. For better or worse, those days have passed.