The Open Source Initiative says Meta is “polluting” the term open-source by using it for Llama, and plans to publish its open-source AI definition next week
Open Source Initiative chief accuses tech group of ‘polluting’ the term by using it to describe its Llama models
Context & Ripple Effects
Meta had positioned the Llama 3.1 family as a frontier-level open-source release, including its largest model, when it introduced Llama 3.1. OSI had already been developing criteria for applying the term to AI systems, rather than treating model-weight availability as sufficient in its work on an Open Source AI Definition.
The dispute matters because Llama is a prominent test case for whether AI developers can market systems as open source while retaining limits around commercial use or underlying development materials. The planned definition moves that argument from branding rhetoric toward a stated standard.
First-order effects
- OSI’s forthcoming definition puts Meta’s use of “open source” for Llama under an explicit criteria-based challenge, raising the reputational stakes of that label for Meta.
- Developers and enterprises evaluating Llama gain a clearer basis for separating access to a model from the broader rights and materials implied by an open-source claim.
Second-order effects
- Other AI model providers using open-source language will face pressure to explain which rights, training-data disclosures, and usage terms their releases provide rather than relying on weight availability alone.
- The disagreement can make licensing and provenance a more visible selection factor for organizations choosing between Llama and alternative models; Meta’s later rejection of the standard underscores that the label will remain contested as Meta disputes OSI’s criteria.
Third-order effects
- If OSI’s definition becomes a common reference point, AI openness is likely to be assessed as a bundle of governance, reproducibility, and use rights—not a binary label attached to downloadable models.
- The market may settle into distinct categories for fully open systems and commercially constrained releases, though adoption will depend on whether major labs, developers, and buyers use OSI’s standard in practice.
The trend: AI model releases are shifting from broad “open-source” branding toward more formal scrutiny of the rights, inputs, and governance that openness requires.