[Introduction] The sanctions clauses of NeurIPS have triggered a powerful backlash! The official’s transfer of “banning first and then consulting lawyers” has been ridiculed, and the China Association for Science and Technology has taken vital motion to institutionally cut back its status to zero. As prime students corresponding to Tsinghua professors collectively withdraw from the overview course of, this storm is just not solely a troublesome counter – assault but additionally fully tears off the fig leaf of the so – known as “top AI conference, in name only”.
NeurIPS is in massive bother now!
The discriminatory coverage below the pretext of “compliance” has sparked robust dissatisfaction and protests within the tutorial circles in China and world wide.
On March 23, the sanctions clauses have been quietly added to the NeurIPS 2026 handbook.
On March 25, the Chinese tutorial circle was in an uproar, and the CCF issued a press release in a single day.
On the afternoon of March 26, after three days of silence, the NeurIPS official lastly posted a press release on X.
The entire assertion is just not lengthy, and the core consists of two paragraphs:
The first paragraph says that “NeurIPS is an inclusive community committed to free scientific exchange.”
The second paragraph says that “the current controversy is not about scientific or academic freedom, but about the legal requirements that the NeurIPS Foundation must comply with, and we are actively consulting lawyers.”
This response was supposed to douse the flames, however as an alternative, the fireplace burned much more fiercely.
The China Association for Science and Technology Steps In: NeurIPS’s Prestige Reduced to Zero
Less than 24 hours after NeurIPS issued its response, the China Association for Science and Technology stepped in instantly –
From now on, cease accepting functions from students for funding to attend the 2026 NeurIPS convention. All related functions will likely be redirected to home tutorial conferences or different worldwide conferences that “respect the rights and interests of Chinese scholars and adhere to the principle of open cooperation.”
An much more extreme measure is that: Papers included on this yr’s NeurIPS is not going to be acknowledged for any tasks utilized to the China Association for Science and Technology as consultant works.
In different phrases, the “prestige” of NeurIPS 2026 within the Chinese tutorial analysis system has been institutionally lowered to zero in a single day.
The Self – Contradiction in a Tweet: “Ban First, Then Ask”
The response of the neighborhood was virtually instantaneous.
The most upvoted remark within the remark part instantly criticized: “So you issued the ban first and then went to consult lawyers? An interesting workflow: ban first, then ask if it’s legal.”
More feedback poured in.
Some folks stated that it was ironic to name a “community that excludes researchers from nearly a thousand institutions” an “inclusive community”.
Some folks urged that NeurIPS transfer its registration location to a impartial nation, “just like the Eclipse Foundation left the United States.”
A researcher stated extra forcefully: “When you enforce a rule that directly determines who can participate in research, it has already become an issue of academic freedom. You can’t separate ‘legal compliance’ from its consequences.”
With 100,000 views, the feedback have been virtually all essential.
NeurIPS tried to distance itself through the use of the phrases “legal compliance”, however the neighborhood did not purchase it.
The Loss of Reviewers Is More Fatal Than the Loss of Submissions
What actually harm NeurIPS was what occurred subsequent – students started to decline invites in massive numbers.
Professor Cui Peng from Tsinghua University instantly declined the invitation to be the Senior Area Chair for NeurIPS 2026 and left a message in his reply:
I’m reluctant to serve for a tutorial convention with such political bias.
(I’m not keen to serve a tutorial convention with such political bias.)
It needs to be famous that the SAC is a core function second solely to the Program Chair within the NeurIPS overview system. Each SAC is accountable for managing the work of about 10 Area Chairs. This stage of refusal is a considerable blow.
Assistant Professor Tan Zhi Xuan from the National University of Singapore additionally publicly posted his letter of withdrawal from the overview on X and added, “If you also oppose this policy, you can directly use my template.”
This publish has now acquired practically 100,000 views.
The trade can be withdrawing.
It is reported that many researchers from main home firms have resigned from their senior positions at NeurIPS 2026.
AI researcher Wang Cunxiang introduced on X that he would decline to be a reviewer, stating bluntly: “I am not willing to serve a highly politicized and racist organization.”
When a tutorial convention begins to lose its reviewers, the issue isn’t just the variety of submissions.
NeurIPS depends on hundreds of volunteer reviewers for its regular operation yearly, and these persons are unpaid volunteers.
The solely driving power for them is their sense of identification with the educational neighborhood.
When this sense of identification is shattered by the sanctions clauses, the lack of reviewers could also be extra deadly than the lack of submissions.
Time Is Running Out for NeurIPS
There are two paths in entrance of NeurIPS now.
One is to observe the instance of IEEE in 2019. That yr, below related stress, IEEE banned Huawei researchers from reviewing papers. After the CCF initiated a boycott, IEEE withdrew the ban inside per week.
The different is to maintain out and bear the lengthy – time period penalties of being faraway from the CCF advisable checklist, a pointy drop within the variety of submissions, and a scarcity of reviewers. Don’t neglect, many main Chinese firms have been the principle sponsors of NeurIPS 2025, and none of them has acknowledged whether or not they’ll renew their sponsorship.
A much bigger query mark is: Will ICML and ICLR observe swimsuit?
The working entities behind the three prime conferences are all registered within the United States and face precisely the identical authorized constraints. NeurIPS was simply the primary to present its hand.
The submission deadline for NeurIPS 2026 is May 6.
They do not have a lot time left to decide.
Is the Top Conference Dead?
The storm of sanctions continues to be raging, however a deeper rift has emerged in the identical week – the overview system of prime conferences is collapsing from inside.
ICML 2026 has simply introduced the overview outcomes, and there are scathing criticisms all around the web.
Ravid Shwartz Ziv, a researcher at Meta, stated that the standard of the opinions was so poor that he “would rather the comments were generated by AI.”
Some reviewers connected the fallacious overview feedback to papers, some requested for experiments that have been already within the appendix, and some fabricated non – existent defects in papers.
ICML acquired about 24,000 submissions this yr, and there have been clearly not sufficient reviewers. Quality management has failed.
More paradoxically, information exhibits that the scores given by AI reviewers are usually increased than these by human reviewers – machines are literally extra “serious” than human reviewers.
Assistant Professor Guanya Shi from CMU posted a protracted thread on X, which has acquired greater than 90,000 views.
He stated that he was “really tired of writing” rebuttal letters towards overview feedback corresponding to “lack of innovation”.
The most typical remark from reviewers is: “This paper simply combines A, B, and C, and has limited innovation.”
Guanya Shi listed eight actual values of A + B + C sort papers in a single breath and refuted them one after the other –
Most papers within the discipline of robotics are “convex combinations” of current concepts. The actual contribution lies in what new capabilities the mix unlocks, what non – trivial interplay mechanisms it reveals, and what previous concepts which have been uncared for for years are revitalized.
Jim Fan, the AI director and distinguished analysis scientist at NVIDIA, instantly lashed out below this publish:
I do not care about convention paper opinions anymore. At the present eve of AGI, it is meaningless.
As extra and extra prime researchers begin to publish their papers instantly on arXiv, share their outcomes on X, and push their code to GitHub, the function of prime conferences as tutorial gatekeepers is being bypassed.
The sanctions have accelerated this course of, however even with out sanctions, this rift would have been uncovered in the end.
Perhaps what NeurIPS needs to be most fearful about is just not the sanctions or the boycott, however that extra and extra prime researchers are beginning to assume that it does not matter whether or not there’s such a platform as you.
Reference supplies:
https://x.com/NeurIPSConf/standing/2037066494983426374
https://x.com/DrJimFan/standing/2036782363318059461
https://x.com/xuanalogue/standing/2036804752265372142
https://x.com/CunxiangWang/standing/2036759558996300015
https://x.com/GuanyaShi/standing/2036675367117725951
This article is from the WeChat official account “New Intelligence Yuan”, edited by: Hao Kun, Allen. Republished by 36Kr with permission.