The Open Source Initiative (OSI) has unveiled a definition framework to evaluate whether AI systems can be classified as open-source.
The announcement of the first Open Source AI Definition (OSAID) was made at All Things Open and marks the culmination of a comprehensive global effort spanning multiple years of research, international workshops, and a year-long community design process.
The OSI – widely recognised as the definitive authority on open-source definitions by individuals, organisations, and government bodies worldwide – developed the framework through extensive collaboration with industry stakeholders. This framework defines what open-source AI means, insisting that the same open-source requirements apply whether to a fully functional AI system, a model, weights and parameters, or other structural elements.
An open-source AI system must be made available under terms that grant four essential freedoms:
- Use the system for any purpose and without having to ask for permission.
- Study how the system works and inspect its components.
- Modify the system for any purpose, including to change its output.
- Share the system for others to use with or without modifications, for any purpose.
These freedoms apply both to a fully functional system and to discrete elements of a system. A precondition to exercising these freedoms is having access to the preferred form to make modifications to the system, which includes detailed data information, complete source code, and model parameters.
“The co-design process that led to version 1.0 of the Open Source AI Definition was well-developed, thorough, inclusive, and fair,” said Carlo Piana, OSI board chair. “The board is confident that the process has resulted in a definition that meets the standards of open-source as defined in the open-source definition and the four essential freedoms.”
One of the framework’s most significant requirements is the mandate for open-source models to provide sufficient information about their training data, ensuring that “a skilled person can recreate a substantially equivalent system using the same or similar data,” according to Ayah Bdeir, who leads AI strategy at Mozilla.
Bdeir acknowledged that whilst this approach might not be perfect, it represents a practical compromise between ideological purity and real-world implementation. She suggested that demanding an unrealistically high standard could prove counterproductive to the initiative’s goals.
The Digital Public Goods Alliance (DPGA) has expressed support for the OSI’s leadership in defining open-source AI. Liv Marte Nordhaug, CEO of the DPGA secretariat, confirmed that her organisation will incorporate this foundational work into updates to their Digital Public Goods Standard for AI applications.
EleutherAI Institute, known for its non-profit work in AI development, has also endorsed the definition.
“The Open Source AI Definition is a necessary step towards promoting the benefits of open-source principles in the field of AI,” stated Stella Biderman, Executive Director of the EleutherAI Institute. “We believe that this definition supports the needs of independent machine learning researchers and promotes greater transparency among the largest AI developers.”
The definition highlights the importance of including data information and code when sharing open-source models and weights. These requirements ensure transparency and the ability to modify the AI system.
OSI Executive Director Stefano Maffulli acknowledged the challenges faced during the development process, noting that despite occasional heated exchanges and differing opinions, the final result aligned with the project’s initial objectives.
“This is a starting point for a continued effort to engage with the communities to improve the definition over time,” he stated.
The OSAID does not require a specific legal mechanism for assuring that model parameters are freely available to all, though it may involve licences or legal instruments. This aspect is expected to become clearer over time as the legal system addresses these open-source AI systems.
See also: President Biden issues first National Security Memorandum on AI
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.
Explore other upcoming enterprise technology events and webinars powered by TechForge here.