On 23–24 September 2026, the Special Interest Group on Artificial Intelligence (SIG-AI) held its eighth meeting in Geneva. With a total of 75 registrations, the event brought together 25 in-person participants, alongside remote attendees, enabling physics experts and NREN specialists to collaborate and discuss the future of AI in our community.
CERN and the AI Ecosystem
The meeting opened with a warm welcome from Eric Grancher (CERN), welcoming and setting the tone for the day. Alberto Di Meglio (CERN) kicked things off with a look at how AI is woven into the organisation’s work today and its plans for the future. Matilde Costa (CERN) then introduced CAFEIN®, CERN’s federated AI platform built for secure, scalable, and collaborative intelligence, offering a glimpse at how computing environments support federated learning in a real scenario. The session closed with Carmen Misa Moreira and Marian Babik (CERN) walking through use cases from the Worldwide LHC Computing Grid (WLCG). This included anomaly detection using AI and ways to improve network visibility with AI, showing how Scitags, perfSONAR, MetraNOVA, and other data collection and data sharing initiatives can benefit from AI and can improve the work of the network people, especially in supporting the understanding of the behaviour of special users like the LHC community.
Scientific Computing and AI Infrastructure
Two talks zoomed in on AI infrastructures and scientific computing. CERIT-SC shared how e-INFRA CZ is rolling out AI inference services and its growing usage numbers. Joining remotely, Hrachya Astsatryan (ASNET-AM) offered a complementary perspective on scientific computing and AI, rounding out the picture of how national e-infrastructures are adapting to support AI workloads.
AI Platforms, Federated Services and Digital Identity
Several presentations focused on how NRENs are building, sharing and federating AI services for their communities. Olivier Wong (RENATER / Université de Rennes) updated us about the ILaaS Federation, while Erik Scott Hognestad (Sikt) gave an update on their Sikt AI platform and how CSC is joining efforts to use and disseminate the services, demonstrating a strong example of NRENs collaborating to disseminate an AI platform across different countries. Esther Ruiz Ben (DFN) presented the potential use of AI for digital wallets.
Open Source, Licensing, and Compliance
Day two opened with Luciana Piccoli (Jisc) discussing AI licensing and how Jisc is dealing with the specifics of AI services procurement. Jakob Tendel (DFN) explored how OCRE can serve as a vehicle for AI adoption across European research, and Marcin Wolski (PCSS) took a deep dive into moving from SBOMs to compliance AI for open source projects. Together, these sessions gave the audience a clear view of the legal and governance groundwork required for responsible AI adoption.
Community Voices: the SIG-AI Open Mic session
Day two also introduced something new to the SIG-AI meeting format: the Community Open Mic. Created to open the floor to community voices and lightning talks, it allowed anyone to jump in with a quick update, a question for the room, or a topic that didn’t make it onto the formal agenda. Haroun Mangal (Uni. Twente/GÉANT) called with a request for DDoS Netflow data to share across the community to support the development and testing of AI for anomaly detection. Simon Leinen (SWITCH) gave a tour of AI-related activities happening within Switch, and Jaco de Vroed (SURF) walked us through SURF’s AI marketplace case.
Shaping the Road Ahead
The final agenda item featured Daniela Brauner (GÉANT) presenting draft work on the GÉANT AI Thematic Roadmap, laying out a shared ambition for the community through 2030. The roadmap was presented as a living draft, inviting the SIG-AI community to weigh in and help set the group’s priorities before it’s finalised.
CERN’s visit: Underground and Beyond
No trip to CERN would be complete without seeing the machines behind the mission, and the meeting closed with exactly that. Participants visited the Antimatter Factory (ELENA, Antiproton Decelerator), where CERN produces and decelerates antiprotons to study the fundamental asymmetry between matter and antimatter. They then descended 82.1 metres underground into the ATLAS experiment cavern, standing beside one of the detectors of the Large Hadron Collider (LHC). These detectors are currently being upgraded as part of the High-Luminosity LHC (HiLumi-LHC) project, which will boost the collider’s collision rate. That leap in data volume is exactly where AI comes in: HiLumi-LHC will generate far more collision data, making AI central to how CERN plans to filter, trigger on, and analyse events at the scale the upgrade demands. It was a striking way to end two days of talks on federated AI platforms, network, open-source compliance, and AI roadmaps.
