Session

DescriptionGenerative AI is reshaping how scientific outputs are produced, yet most widely used tools are operated by commercial providers whose practices around processing, generating, and storing user inputs are often unclear. This lack of transparency raises serious concerns for research organizations handling sensitive or regulated data and complicates the responsible use of even non-regulated scientific content. At the same time, many commercially served models are closed-source and subject to frequent, opaque updates, limiting reproducibility and undermining alignment with FAIR principles. As open-weight and open-source models proliferate, HPC infrastructures are emerging as a promising alternative for hosting and providing controlled access to AI within research environments. However, most HPC systems were not designed for continuous, GPU-based inference services, creating technical and operational challenges spanning deployment, scheduling, reliability, user access, and governance. Consequently, institutions are developing ad hoc solutions with limited opportunities to exchange patterns and lessons learned. This minisymposium convenes a panel of institutions actively building such capabilities to compare approaches and discuss best practices, minimal viable solutions, and ideal configurations for serving generative AI on HPC. To broaden participation, we will use a short questionnaire to structure contributions across four dimensions: technical setup, usage policies, documentation practices, and monitoring/oversight mechanisms.