Yale researchers are floating a copyleft framework for generative AI, borrowing the reciprocal-sharing logic of open-source software licenses to govern how models trained on copyrighted or user-generated content must share outputs or terms downstream. For data companies and AI developers, such a regime would mean training-data provenance and licensing terms could travel with a model's outputs, not just its inputs—raising compliance stakes well beyond the current scrape-and-train dispute.
Details on enforcement mechanisms or industry reaction were not specified in the available reporting, according to YaleNews.
Yale researchers propose 'copyleft' rules for generative AI
— YaleNews
