Open Compliance Summit 2024

In the past year, many high-quality AI models have been "open sourced," but under various licenses. Google and Meta, for example, have adopted self-created licenses, while Stable Diffusion moved from the standardized OpenRail-M to a self-created license. In contrast, xAI chose the traditional Apache 2.0 license to "open source" Grok. Open source organizations like the Linux Foundation, the Open Source Initiative, and the Chinese OpenAtom Foundation are working on drafting frameworks, definitions, and standardized licenses to address the rising selection costs caused by non-standard licenses. The complexity and lack of standardization in AI model licenses arise because AI models, as structured data, present more confusing intellectual property issues compared to traditional software, which benefits from the Berne Convention and the WIPO Copyright Treaty (WCT). This session will explore the intellectual property challenges surrounding AI models, leveraging the experience of drafting the OpenAtom Model License. We aim to clarify what rights are actually licensed and outline what a robust license should cover, offering recommendations for future license drafting and standard-setting.