NeurIPS Reverses Controversial Policy After Backlash From Chinese AI Researchers

NeurIPS, the world's top artificial intelligence conference, has found itself at the center of a scandal: a change to its participation policy, announced by the organizers, drew a barrage of criticism from Chinese scientists and was reversed within a few days.

NeurIPS (Conference on Neural Information Processing Systems) brings together tens of thousands of researchers from the world's leading universities and corporations every year. It is considered the gold standard in academic AI: having a paper accepted there carries a weight comparable to publishing in Nature or Science in other fields. Chinese participants, affiliated with Tsinghua, Peking University, Shanghai AI Lab and the country's leading tech companies, make up one of the largest national contingents among authors and reviewers.

Wired's public materials give few details about the specific policy change, but the reaction is telling: the wave of protest spread across X, WeChat and academic forums. Researchers voiced concern that the rules effectively restrict participation by specialists from China, on the basis of affiliation or citizenship rather than scientific merit. Some authors said they were ready to boycott the conference or withdraw submissions they had already filed.

NeurIPS organizers responded quickly, and the decision was reversed within a few days. The committee explained that the change had been introduced with good intentions but was poorly communicated and led to misinterpretations. The official statement stressed that the conference is committed to the principle of open science regardless of participants' national origin.

Still, the incident is symptomatic of a broader trend. US semiconductor export restrictions, the Commerce Department's "blacklists," limits on academic exchange and visa barriers are gradually reshaping the architecture of the global AI community. Conferences find themselves in an ambiguous position: they want to remain neutral scientific venues, but they cannot ignore the regulatory risks facing sponsors and member organizations that operate under US jurisdiction.

The context of the technology race plays its own role. After the success of DeepSeek R1 and V3 in early 2025, it became clear that Chinese labs can build world-class models with significantly less compute. That changed how China is seen: from a "catch-up" player to a full-fledged competitor. Against this backdrop, voices calling for limits on scientific exchange with Chinese institutions, citing dual-use risks, are getting louder.

All of this raises a painful question for the global AI community: can academic science remain truly international when its main achievements directly affect the military and economic power of states? Conferences such as NeurIPS, ICML and ICLR still remain open, but every incident like this tests that consensus.

In the short term, the reversal looks like a victory for common sense. But the underlying contradiction between the principles of open science and geopolitical interests has not gone away. The next incident like this may be resolved differently.