Discover the Non-Profit Working to Define Open Source AI Standards

The concept of open-source AI lacks a universally accepted definition, prompting efforts from one nonprofit organization to establish a comprehensive standard. The Open Source Initiative (OSI) is leading the charge to develop a clear and globally recognized definition for open-source AI, akin to the widely acknowledged Open Source Definition for software.

“AI differs fundamentally from traditional software, challenging all stakeholders to reconsider how open-source principles apply in this field,” stated Stefano Maffulli, OSI’s executive director. “We believe it is essential for everyone to maintain agency and control over technology. Furthermore, markets thrive when transparency fosters collaboration and innovation without barriers.”

Prominent figures in the AI community, such as Yann LeCun, advocate for an open-source future for AI. However, the absence of a clear definition opens the door for debate on what truly qualifies as an open AI system. Key technology companies, including Microsoft, Google, and Meta, have released AI models—Phi-3, Gemma, and Llama 3—under open-source terms. Yet, despite releasing model weights and source code, some of these organizations do not disclose the foundational training data.

Seth Dobrin, founder and CEO of Qantm AI and former chief AI officer at IBM, refers to these practices as “pseudo-open source.” Julia Ferraioli, a machine learning strategist at Amazon, emphasized at a recent industry event that the mere availability of a system does not inherently make it open.

Recognizing the need for clearer standards, OSI has dedicated the last two years to research in collaboration with various stakeholders to establish guidelines for open-source AI. After compiling what it describes as “a massive body of work,” the organization is embarking on a global tour to present its insights and refine its open-source AI definition.

“After gathering diverse perspectives from around the world to define principles suitable for open-source AI systems, we are kicking off a worldwide roadshow to validate the draft of our open-source AI definition,” remarked Maffulli.

This initiative will bring OSI's findings to several international events, including those in Paris, Lagos, Hong Kong, and Argentina. In the U.S., workshops are scheduled for Pittsburgh at PyCon, alongside sessions in New York and Raleigh, North Carolina, allowing for broad engagement and collaboration across the tech community.

Through these efforts, the OSI aims to clarify what constitutes open-source AI, ensuring that the principles guiding this transformative technology are transparent, inclusive, and conducive to innovation.