Researchers at Yale University have proposed a new copyleft licensing framework that would require AI developers using open-source code to disclose model architecture and training data, aiming to keep generative AI transparent and aligned with open-source principles.
Researchers at Yale University’s Digital Ethics Center (DEC) have proposed a new licensing framework designed to ensure that generative AI models built on open-source code remain open and transparent.
Called the Contextual Copyleft AI License (CCAI), the proposal extends traditional open-source copyleft principles to generative AI by treating AI models trained on open-source code as derivative works. Under the framework, developers using open-source software for AI training would be required to publicly disclose their model architecture and training data.
The researchers argue that many AI companies benefit from open-source software without providing equivalent transparency in return, leaving developers with little visibility into how their code is used or whether resulting AI systems remain open.
“Our analysis showed that extending the copyleft concept to generative artificial intelligence has the potential to give open-source software developers meaningful control over how AI developers use their code,” said Grant Shanklin, lead author and de Vries-Sherif Junior Fellow at DEC.
According to the study, the proposed licence would give developers greater control over AI use of their code, encourage the creation of more fully open-source AI models, improve transparency and accountability, and help curb “open washing” by organisations that market proprietary AI systems as open.
“AI companies have benefited from using open-source code, but their resulting models are not really open,” said Claudio Novelli, de Vries-Sherif Associate Research Scientist at DEC.
Published in the International Journal Of Law And Information Technology, the study concludes that CCAI licensing is legally feasible under current copyright law, provided AI training is not considered fair use. The researchers also argue that existing regulations, including European Union AI rules, could help mitigate risks associated with open-source generative AI.
