We finally have an 'official' definition for open source AI | TechCrunch
The Open Source Initiative has released the first official definition of open-source AI, aiming to resolve confusion in a rapidly evolving landscape. To qualify as open source under this new standard, AI models must provide complete transparency regarding their design, training data provenance, and code, enabling users to substantially recreate and build upon the system. This framework seeks to align industry practices with regulatory expectations, particularly in regions like the European Union, by establishing a clear consensus on what constitutes genuine openness versus marketing claims. This definition directly challenges major tech companies that label restrictive models as "open source." By setting strict criteria for accessibility and modification rights, the standard exposes practices that limit usage to large platforms or require expensive enterprise licenses, effectively labeling them as pseudo-open. The initiative aims to leverage community consensus to correct these misrepresentations, ensuring that the term "open source" reflects true democratization of technology rather than serving as a veil for centralized control and limited access. The relevance to open data is profound, as the definition mandates full disclosure of training data sources and processing methods. This transparency requirement forces developers to confront legal and ethical complexities surrounding data scraping and copyright, potentially impacting how datasets are licensed and shared in future iterations. By highlighting the gap between current industry practices and true openness, the definition underscores the critical need for clear, accessible data standards to prevent the entrenchment of power among a few tech giants and to foster genuine innovation in the AI community.
Source: techcrunch.comPublished on 2024-10-29