Open source AI is forcing a legal reckoning: copyright law, which has long governed how open source code can be shared and used, may not work the same way for AI systems.

What happened: McCoy Smith, speaking at an Open Source Initiative webinar, outlined how copyright has been the primary legal foundation for open source licensing, but AI is raising fundamental questions about how copyright applies to training data, the processing that happens inside an AI system, and the output that AI produces.

Why it matters: There is ongoing litigation and debate about whether creators of materials used to train AI have any legal right to object to that training, and whether AI output itself can be copyrighted or patented. The core problem is that the middle step—what happens between input and output inside an AI system—remains legally unresolved, and no clear legal regime yet exists to govern it the way copyright licensing does for traditional open source code.

What to watch: Smith speculated on potential future solutions, including legislative action, contract-based approaches, and public pledges, suggesting that copyright law alone may not be sufficient to keep AI truly open.