# OSI hosts AI Salons to refine Open Source AI Definition (OSAID) standards

> **Key takeaway:** The Open Source Initiative is engaging global stakeholders to address the technical and legal challenges of applying open source principles to AI weights and training data.

**Published:** 2026-11-10T17:58:59+00:00 
**Source:** Open Source Initiative (OSI) 
**Category:** policy-community 
**Canonical URL:** https://fosswire.org/news/osi-hosts-ai-salons-to-refine-open-source-ai-definition-osai.html 

## Summary
The Open Source Initiative (OSI) has launched a series of 'Open Source AI Salons' to facilitate community discussions regarding the Open Source AI Definition (OSAID). Following the 2024 release of OSAID 1.0, these gatherings in Raleigh and Barcelona aim to address the friction between traditional open source software norms and the unique requirements of artificial intelligence.

As the AI ecosystem expands—evidenced by Hugging Face hosting over three million models—the OSI is investigating how to maintain the four fundamental freedoms of use, study, share, and modify within AI. The initiative is specifically targeting the complexities of open-weight models, the emergence of AI-specific licenses like ModelGo and OpenMDW, and the critical need for transparency in training datasets.

## Challenges in applying OSAID to AI models
- **Licensing Discrepancies:** Developers frequently release code under OSI-approved licenses while simultaneously imposing restrictive Acceptable Use Policies (AUP) on model weights, which may violate OSAID principles.
- **Data Transparency Requirements:** A primary technical bottleneck is the OSAID requirement for detailed information regarding training data, a standard that few currently available models meet.
- **Weight vs. Code Heuristics:** Traditional open source heuristics do not translate cleanly to AI, necessitating new consensus on how to define openness when dealing with model weights rather than just source code.
- **Metadata and Provenance:** Discussions have highlighted the need for machine-readable metadata to communicate model provenance, use restrictions, and sharing terms at scale for automated governance.

## Why it matters
Why it matters: As governments introduce new AI regulations, the OSAID provides a framework for developers to ensure their models are truly open rather than just 'open-weight.' This work is critical for software engineers and distributors who need clear, standardized definitions to navigate the legal complexities of integrating AI into open source stacks.