The Open Source Initiative (OSI) has unveiled its official definition of “open” AI, requiring transparency about the data and methods used to train models. This move challenges tech giants like Meta, whose Llama AI model, while publicly available, doesn’t provide access to training data or fully meet OSI’s open-source criteria.
For AI to be considered open, OSI states it must disclose:
- Training data details
- Complete source code
- Training settings and weights
This redefinition reshapes AI standards, aligning with OSI’s principles of freedom to modify and share. The industry now faces pressure to adopt or oppose these values, as the debate over open-source AI intensifies.
