📠Sydney, Australie 📅 6 au 13 décembre 2026
From December 6 to 13, 2026, NeurIPS 2026 will once again bring together the global artificial intelligence community around one of the most influential scientific meetings in the sector. The 2026 edition will take place in Sydney, with two official satellites in Atlanta and Paris, confirming the gradual transformation of the conference into a distributed global platform.
Founded in 1987 in Denver, NeurIPS has established itself as the reference conference in the fields of machine learning, deep learning and artificial intelligence. What was once a specialized academic circle has become a strategic center of gravity for the entire global technology industry. Researchers, hyperscalers, public laboratories, startups, investors and public authorities now meet there to observe the advances that will shape future generations of AI models and infrastructures.
The 2026 edition will be held in Sydney from December 6 to 12, in Atlanta from December 8 to 13 and in Paris from December 9 to 13. For six days, the conference will alternate keynotes, oral presentations, poster sessions, workshops, tutorials, technical competitions and community events. The world’s leading research teams will unveil their latest work on multimodal models, autonomous agents, architecture optimization, generative systems and even intensive computing issues.
But NeurIPS remains above all an extremely rigorous scientific selection machine. Each submitted article is evaluated by several independent experts in a review process considered one of the most demanding in the industry. By 2024, more than 15,000 articles had been submitted, for an acceptance rate of less than 24%. This requirement helps maintain the conference as one of the world’s leading standards for scientific validation in AI.
For several years, NeurIPS has also strengthened its methodological and ethical criteria. Authors must now document the availability of data and code, explain the limits of their work and include reflection on the societal impacts of the research presented. This development reflects the transformation of the sector: AI is no longer only evaluated on its technical performance, but also on its regulatory, environmental and political implications.
The 2026 edition also introduces a finer structuring of scientific contributions. The conference now distinguishes several categories of work: general research, theory, use-oriented research, exploratory concepts with high potential, negative results, datasets and position papers. This segmentation reflects a profound evolution in the sector, where scientific value no longer rests solely on the raw performance of a model, but also on the quality of evaluations, methodological robustness or the capacity to question certain dominant paradigms.
Among the examples highlighted by NeurIPS are works that have become structuring for the entire industry, such as “Segment Anything†, the “Neural Tangent Kernel†, the “Capsule Networks†or even several research studies on the fundamental limits of algorithmic fairness and the generalization of neural networks. This ability to detect future paradigms very early explains why large technology groups are now monitoring the conference with as much attention as an industrial summit.
But the 2026 edition also opens in an unusually sensitive context for the scientific community. In recent days, NeurIPS has had to manage a controversy linked to its policy of compliance with international sanctions. In the official conference guide, a link referred to an American government tool covering a much wider scope of restrictions than that actually applicable to NeurIPS. Quickly relayed within the AI community, this reference raised concerns about a possible tightening of participation conditions for certain researchers and international institutions.
Faced with the reactions, the organization published an unusually direct press release. NeurIPS acknowledges a communication error between the conference foundation and its legal team, while affirming that there was never any question of extending the restrictions beyond the regulatory obligations imposed by the American authorities. The organization reaffirmed that the conference remained open to “all institutions and individuals complying with applicable obligations.”
Beyond the incident itself, this sequence illustrates the new geopolitical reality of artificial intelligence. Major scientific conferences are no longer just neutral academic spaces. They become strategic infrastructures exposed to international tensions around semiconductors, foundation models, high performance computing and technological sovereignty policies.
This development profoundly modifies the role of NeurIPS. The conference is no longer just a place where scientific papers are presented. It now acts as a global observatory of technological power relations. Large companies recruit the most sought-after researchers, investors identify future industrial standards, while governments monitor advances likely to have economic, military or strategic implications.
For AI startups, obtaining a publication at NeurIPS remains a considerable accelerator of scientific and industrial credibility. For academic laboratories, the conference remains one of the main points of access to international visibility. But in a context marked by the computing war, restrictions on chip exports and the growing geopolitical fragmentation of AI, NeurIPS now appears to be much more than a scientific conference: a space where the future global balances of artificial intelligence are taking shape.
