The ARTIFACT Strategy: Federated AI for Particle Accelerator Research Infrastructures

Adnan Ghribi (GANIL – DOD GPA)

Particle accelerators are among the most complex scientific instruments ever built, yet the adoption of artificial intelligence across these facilities remains fragmented, facility-specific, and difficult to sustain beyond individual projects. The ARTIFACT network — 41 partner institutions across 14 countries — was established to change this, by coordinating a concerted, open-science strategy for AI in accelerator science and its user communities.

This seminar presents the ARTIFACT strategic framework, structured around six pillars from FAIR data infrastructure to AI policy, and introduces the two recently funded European projects that are now at the stage of grant agreement preparation. TwinRISE builds a trusted, cross-disciplinary platform for AI-powered digital twins, coupling physics simulations with federated machine learning across accelerator, medical, and energy facilities. iRIS deploys AI-driven sustainability technologies at major European research infrastructures — including HL-LHC, FCC, FAIR, and GANIL — through eco-design modeling, energy optimization, and shared AI model exchange.

Beyond these two projects, I will discuss the open strategic gaps: training the next generation of researchers, standardizing AI deployment within control systems, bridging machine intelligence with user science, and building governance frameworks that outlast individual funding cycles. Perspectives for upcoming initiatives — including a RAISE doctoral network and a CNRS International Research Network — will be outlined.