FROM SOFTWARE TO MODELS: RETHINKING OPEN SOURCE IN THE AGE OF GENERATIVE AI

Authors

Dana Beldiman

Abstract

Open-source software (OSS), a form of distributed and collaborative software development rooted in the ethos of early computer science, has evolved into an essential component of modern digital infrastructure. This article asks whether open source will assume a similarly transformative role within the emerging technological paradigm of generative artificial intelligence (AI).

Generative AI differs fundamentally from software: its constituent artifacts, architectures, generative capacities, and deployment contexts are distinct and substantially more complex. Despite these differences, foundation models continue to be released under licenses originally designed for software, which may not constitute a valid transfer of rights in the generative AI context. Closer examination raises serious doubts regarding the copyrightability of trained models and, consequently, regarding the validity and enforceability of conventional open-source licensing schemes when applied to AI systems.

Why, then, are an increasing number of large language models (LLMs) released under open-source licenses? This article advances two explanations. First, collaborative innovation can produce accelerated, high-quality, and cost-effective outcomes that compare favorably with proprietary development. Second, open-source releases serve as a strategic instrument in the competitive race among foundation model deployers. By releasing foundation models “into the commons,” deployers capture cumulative innovation and network effects generated by the broader community, while contributors gain free access to a sophisticated technological base.

This strategy-driven logic privileges commercial considerations over normative commitments to openness. Nevertheless, compared to fully proprietary development, open-source releases of LLMs retain a degree of functional openness that preserves space for innovation and contributes in some manner to the democratization of the AI ecosystem.

Recommended Citation

Dana Beldiman, FROM SOFTWARE TO MODELS: RETHINKING OPEN SOURCE IN THE AGE OF GENERATIVE AI, 21 Wash. J. L. Tech. & Arts (2026).
Available at: https://digitalcommons.law.uw.edu/wjlta/vol21/iss2/6