“Developers involved in projects that move to a neutral foundation … benefit from the separation of business … and technology in its governance,” Bryant says. “Business leadership and investment in non-technical concerns – legal, funding, marketing, community events, and so forth – are addressed, while developers are able to focus on collectively driving the technology roadmap and releasing code.”
While the open source AI definition remains controversial, and very few AI projects fully qualify as open source by the strict requirements of the Open Source Initiative (OSI) AI definition, AI remains built on a foundation of open source software. The debate over open weights, data and training code will continue, but even the most proprietary large language models (LLMs) couldn’t exist without open source programs.
As lead author Shayne Longpr and team pointed out, “The Open Source Initiative defines open source AI models as those which have open model weights, but also ‘sufficiently detailed information about their [training] data’,” adding, “Without training data disclosure, a released model is considered ‘open weight’ rather than ‘open source’.”
After a number of notable moves to make software proprietary, some companies are insisting on Open Source Initiative-approved software licenses.
“We had some trepidation about that. It turns out, our fears were maybe a little bit unfounded, not entirely unfounded, but a little unfounded. And meanwhile, we knew that there were some customers who really didn’t want to look at our technology unless it was open source, capital ‘O,’ capital ‘S,’ with an OSI-approved license,” Merrick said.
At the Open Source Summit in Amsterdam, I sat down with Maffulli to talk about the current state of the discussion. He noted that not only has the conversation started but that the definition has become a tool for the OSI to engage with politicians, including the European Commission, where the AI Act, for example, will go into full effect in August 2026.
The Open Source Initiative (OSI), a nonprofit that sets the foundation for the open source software ecosystem, did the hard work to define open source decades ago. It identified 10 criteria that include free redistribution, integrity of the author’s source code and no discrimination against persons and groups, among others. These criteria are the guarantees that allow companies to use OSS without calling their legal department every time a developer installs a package.
For their contribution to Maintainer Month, the OSI has collected nearly four dozen stories from maintainers to create a moment of visibility, publishing them all as a curated collection at OpenSource.org
Last summer, the Open Source Initiative and Open Future convened a group of experts to explore this challenge and propose a path forward. A recently released report from the convening, “Data Governance in Open Source AI” argues that collective action is needed to release more data and improve data governance to balance open sharing with responsible release.
The Open Source Initiative, after years of planning, in October 2024 introduced its initial definition of open source AI, which addresses four different kinds of data and requires those building and sponsoring AI technology to share what data they can as well as the model’s parameters and the source code used to train and run the system.
The Open Source Initiative (OSI) has officially released version 1.0 of its Open Source AI Definition (OSAID) on Oct. 28 at the 2024 All Things Open conference this week. It’s been a long slow journey to this significant milestone in the effort to establish a clear standard for open source artificial intelligence (AI).
The Open Source Initiative’s Release Candidate 1 identifies four categories of data and demands sharing of data, source code and model parameters.
The Open Source Initiative has been working on defining open source AI for the past two years, and they should release their definition in November,” Columbro said. “OSI’s approach will focus on the principles, providing a binary definition of open source AI — either it is or it isn’t.
