A community-led project creating an open-source corpus of AI training data with provenance information needed to understand and reproduce AI models.
OpenWALDO, an open-source artificial intelligence project sponsored by Ctrl IQ Inc., has been launched under the leadership of Gregory Kurtzer, founder of Rocky Linux, CentOS and Apptainer. The project aims to create a community-led, open-source corpus of AI training data, providing a shared foundation that AI builders can use rather than duplicating foundational work.
Open-weight models offer the capability to download and execute AI models on customer hardware or in cloud computing platforms; however, they do not necessarily include the code, methods, recipes and training data used to create them. To bridge this gap and ensure transparency about what is being used to generate an AI model, OpenWALDO aims to publish this information openly.
An AI Bill of Materials containing information such as object references, documents, tokens, inventory and licences as evidence of the content used to build an AI model. These records are intended to make training data attributable, reviewable and correctable in public, allowing developers to examine the provenance of the material used to train a model. This can also help verify the licensing and copyright status of the sources before they are used in model development.
For collaboration, OpenWALDO is intended to support collaboration among AI communities by providing a shared space to store proven practices, develop verified baselines, conduct public experiments and create auditable development chains that can be evaluated collectively. The initiative is open to individuals, hobbyists, researchers, organizations and corporations for contributing to building a common foundation for AI development.
The initiative emphasizes provenance and transparency as key elements of open AI development and it applies open-source principles to AI training data and model development. The objective behind the initiative is to save duplicate efforts and develop a resilient public infrastructure for developing AI models.
