Máirín Duffy, Red Hat | Red Hat Summit 2025
summary
Argues that open source matters less as a label and more as artifact-level transparency: models are a mix of data, code and parameters across pretraining, fine-tuning and inference, so licenses must be applied per artifact not monolithically. Demonstrates use of OSI open source AI definition and the Linux Foundation AI MOF for classification, highlights Podman Desktop AI Lab and InstructLab as practical toolchains, recommends indemnified models like IBM’s Granite for product use, notes standing-room-only interest and gives concrete legal questions (redistribution, commercial inclusion, pretraining IP) to ask before deploying.
AI-written summary · about
