The Case for Contextual Copyleft: Licensing Open Source Training Data and Generative AI

Open source AI needs new rules: Traditional copyleft licenses (which require derivative works to remain open source) don't adequately address generative AI, creating a gap where companies can use open source code to train AI models without sharing their improvements—a practice the paper calls "openwashing."

Contextual Copyleft AI license could level the playing field: The proposed CCAI license would require that AI models trained on open source data remain open source themselves, giving developers more control over their work and encouraging companies to contribute back to the open source community rather than keeping AI improvements proprietary.

Legal and practical challenges require careful implementation: While the paper argues CCAI licenses are legally feasible under current copyright law, they work best alongside government regulations that address AI-specific risks like misuse and safety concerns—licensing alone isn't sufficient.

Balancing innovation with protection is critical: The paper acknowledges that stricter copyleft requirements for AI could create friction in AI development, so any new licensing approach must be carefully designed to encourage open source AI innovation while preventing harmful uses.

The stakes are higher with AI than traditional software: Because generative AI models pose unique risks (including potential for direct misuse), the paper argues that open source AI requires stronger safeguards than traditional open source software, making complementary policy and regulatory frameworks essential.