AI is redefining what it means to be Open Source. While the Open Source Definition (OSD) has safeguarded software freedom for over 25 years, it was built for an era of code, not data-driven AI. Today’s foundation models rely on vast, often proprietary datasets, raising critical questions about openness, transparency, and control. If Open Source is to thrive in this new era, we must reimagine its possibilities—not just in code, but in data and AI itself.

This BoF will explore Open Source 2.0, a vision for adapting Open Source to the AI-driven future. We’ll discuss the Open Source Alliance (OSA), an official AI Convergence Challenge at the AI Action Summit (Paris, Feb 10-11, 2025), which seeks to build global consensus on Open Source AI . We will also examine efforts to extend the Open Source Definition to explicitly include data, ensuring that AI systems respect the fundamental freedoms to use, study, modify, and share—not only for code, but for the datasets that power them.

At the same time, there is growing recognition of Open Weight models—AI systems where model weights are published but essential components like training data are not disclosed. To bring clarity, the Open Weight Definition (OWD) has been introduced as a distinct framework that acknowledges these models while differentiating them from Open Source AI . The OSI’s Open Source AI Definition (OSAID) and the EU AI Act both describe Open Weight models more so than Open Source AI in the traditional sense, reflecting the need for clearer language in AI regulation and governance.

Join us to debate the future of Open Source AI: How do we ensure its integrity? How do we define openness in an age where data is the key input? And how can we prevent the dilution of Open Source principles while fostering real innovation?